2202 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 45, NO. 7, JULY 2007

A Compact 183-GHz Radiometer for WaterVapor and Liquid Water Sensing

Andrew L. Pazmany, Member, IEEE

Abstract—ProSensing Inc. has developed a G-band (183 GHz)water Vapor Radiometer (GVR) for long-term unattended mea-surements of low concentrations of atmospheric water vaporand liquid water. Precipitable water vapor (PWV) and liquidwater path (LWP) are estimated from zenith brightness temper-atures measured using four double-sideband receiver channels,which are centered at 183.31 ± 1, 183.31 ± 3, 183.31 ± 7, and183.31 ± 14 GHz. A prototype ground-based version of the in-strument was deployed at the Department of Energy AtmosphericRadiation Measurement program’s North Slope of Alaska sitenear Barrow, AK, in April 2005, where it collected data contin-uously for one year. This paper presents design details, laboratorytest results, and examples of retrieved PWV and LWP from mea-sured brightness temperature data.

Index Terms—Millimeter-wave radiometry, precipitable watervapor (PWV) and liquid water path (LWP) retrieval, remotesensing.

I. INTRODUCTION

MOST ground-based atmospheric water vapor radiome-ters are designed to measure blackbody radiation near

the 22-GHz water vapor absorption line, despite the fact thatthe 183-GHz line is about 50 times more sensitive to changesin precipitable water vapor (PWV) and more than ten timesmore sensitive to liquid water path (LWP) [1]. This preferenceis primarily because the center of the 183-GHz line saturatesat a relatively low 2-mm PWV, making this frequency unsuit-able for general-purpose year-round observations. Furthermore,183-GHz instruments have been considerably more difficultand expensive to build due to the historic lack of off-the-shelfmicrowave components above 100 GHz.

In arid regions, including high latitudes, deserts, or abovethe atmospheric boundary layer, PWV measurement accu-racy of a few tenth of a millimeter is required to monitorchanges in humidity. In dry (few millimeters PWV) conditions,183-GHz brightness temperature changes by more than 20 K,as shown in Fig. 1, for each millimeter change in total watervapor; thus, an instrument with a 1-K radiometric measurementprecision can detect 0.05-mm change in the vapor column.This same measurement resolution would require a precisionof less than 0.02 K with a 22-GHz radiometer, which is alevel of precision that only a handful of radiometers havebeen able to approach, requiring an expensive developmenteffort [2]. On the other hand, the 1 K precision needed at

Manuscript received May 30, 2006; revised September 26, 2006. This workwas supported by the Department of Energy under Phase II Small BusinessInnovation Research Contract DE-FG02-02ER83440.

The author is with ProSensing Inc., Amherst, MA 01002 USA (e-mail:pazmany@prosensing.com).

Digital Object Identifier 10.1109/TGRS.2006.888104

Fig. 1. Sensitivity of brightness temperature to changes in PWV(∆Tb/∆PWV) as a function of frequency and total vapor near the 183.31-GHzvapor line. In very dry conditions, Tb measurements close to the vapor lineare most sensitive to water vapor, but with increasing PWV, the optimumfrequency shifts away from the line (calculated using the radiative transfermodels compiled in [6]).

183 GHz is a routine radiometer design goal. Nevertheless,at millimeter wavelengths, front-end losses, effects of radomereflections, and complexity of incorporating stable calibrationloads in the radiometer present significant design challenges.Several recent advances, however, make the development of a183-GHz radiometer more practical today. These include thecommercial availability of subharmonically pumped Schottkymixers, a high-beam-efficiency conical feedhorn developedfor a satellite radiometer, and a temperature-stable packag-ing technique developed by Tanner for an ultrastable K-bandradiometer [2].

The majority of 183-GHz radiometers constructed to datewere custom designed and developed for spaceborne operation[e.g., one channel of the NOAA-15 Advanced MicrowaveSounding Unit (AMSU)] or were built to operate as part ofground-based millimeter-wave interferometric telescopes suchas the James Clark Maxwell Telescope [3] and the AtacamaLarge Millimeter Array telescope to track optical path varia-tions. Two research ground-based instruments were also con-structed by the Environmental Technology Laboratory and byNASA and were both operated in March 1999 in Barrow, AK, tocompare the ability of 22- and 183-GHz radiometers to measurewater vapor in winter arctic conditions [1]. More recently, theground-based scanning radiometer [4] participated in a latewinter Intensive Observation Period (IOP) at the Department of

0196-2892/$25.00 © 2007 IEEE

PAZMANY: COMPACT 183-GHz RADIOMETER FOR WATER VAPOR AND LIQUID WATER SENSING 2203

Fig. 2. GVR simplified component-level block diagram.

Energy (DOE) North Slope of Alaska Atmospheric Measure-ment (ARM) program site near Barrow, AK, in 2004.

This paper describes a compact, turn-key, four-channelG-band (183-GHz, 3-mm wavelength) water Vapor Radiometer(GVR), which is designed for long-term unattended operationon the ground.

II. GVR DESCRIPTION

A simplified block diagram of the GVR receiver is shownin Fig. 2. The downwelling atmospheric radiation is capturedby a 10-cm-diameter, 1.7◦ 3-dB (half-power relative to maxi-mum gain) beamwidth, 90◦ parabolic metal mirror and focusedto a corrugated feedhorn. A subharmonically pumped mixer,using a six times multiplied 15.276-GHz dielectric resonantoscillator local oscillator (LO) signal, downconverts the upperand lower sidebands to baseband, where a broad-band low-noise amplifier (LNA noise figure < 2.3 dB) increases the noisesignal power and sets the receiver noise temperature. A broad-band power splitter divides the amplified signal between fourchannels before filtering. The center frequency and bandwidthof the filters are 1/0.5, 3/1, 7/1.4, and 14/2 GHz, respectively.The band-limited noise signals are square-law detected, con-verted to a transistor–transistor logic pulse train using highlylinear Analog Devices AD650 voltage-to-frequency converters(20-ppm nonlinearity and less than 0.3-ms response time to astep input as configured), and frequency counted using a field-programmable gate array (FPGA) processor. This frequencycounting effectively integrates and measures the square-lawdetector voltage or, equivalently, the noise power. The measurednoise power from the four channels together with the instrumenttemperature readings are time stamped and transmitted to a datalogger personal computer (PC) via an RS422 serial bus.

Since neither noise sources nor low-loss fast switches arereadily available at G-band, GVR is operated as a total powerradiometer, and external (to the antenna horn) hot and warmcalibration absorbers are periodically observed to track changesin the receiver gain and offset. The metal mirror of the groundGVR is rotated with a stepper motor to point the radiometerantenna beam to the calibration loads. The hot and warmcalibration absorbers were constructed using the Firam-160absorber, with the hot load packed in an insulated box coveredwith a 1-mil Mylar window (∼0.03-dB loss factor) tilted by 4◦.The Mylar window is tilted by about twice the antenna 3-dBbeamwidth to minimize the reflection of the receiver emittedradiation back to the receiver. The hot load is convection heatedto a uniform temperature of 343 ±0.5 K, whereas the warm load

is left to soak to the temperature of the outer enclosure, whichis heated with a proportional–integral–derivative controller to293 K. The temperature of the absorbers is monitored withresistance temperature detectors, and the readings are recordedalong with the receiver noise power data. The temperaturesensors are calibrated with a laboratory calibration certifiedHart Scientific 1502A meter and 5623A probe to better than0.1 K accuracy.

Special care was also taken to maximize the antenna beamefficiency. The 10-cm-diameter, 90◦ optical quality metal col-lector mirror has a root-mean-square (rms) surface roughnessof less than 175 Å (175 × 10−10 m), which is a negligible0.001% of the radiometer wavelength. Since there is no feed-horn or subreflector blockage with a 90◦ mirror, the only othercritical factor for maximizing beam efficiency was the mirrorillumination. The GVR feed is a copy of the space-qualifiedAMSU-B satellite instrument feed, with a well-characterized,low-sidelobe pattern (23◦ 3 dB beamwidth, −30-dB first side-lobe). When placed at the mirror focal point, 99.5% of thefeed pattern is intercepted by the mirror surface, and the 3-dBbeamwidth of the feed illuminates roughly half of the mirror’soverall diameter. The cost of this underillumination is that theGVR antenna 3-dB beamwidth is broadened to 1.7◦ comparedto about 1.1◦ beamwidth of a same-sized antenna designed tomaximize gain.

This mechanical method of calibration can only be repeateda few times a minute; thus, the receiver gain must remain stableon this timescale. For each Kelvin change in physical temper-ature of the GVR amplifiers (component plate), the receivergain changes by about 0.1 dB or, equivalently, an 80-K changein the perceived scene brightness temperature. Consequently,temperature stabilizing the receiver components, particularlythe amplifiers, is essential.

To stabilize the temperature of the receiver components, thepackaging technique developed by Tanner [2] was employed,consisting of an insulated cold plate in a box, in a temperature-controlled enclosure. The resulting component plate temper-ature stability limited receiver gain rate-of-change to below10 dB/h (100 K/h equivalent scene temperature drift rate),making the calibration cycle rate of a few times a minutesufficient to achieve a sub-Kelvin measurement precision.

The filter-bank-type receiver was chosen over a variableLO-type design to increase data rate and to keep the high-frequency portion of the instrument as simple as possible. Avariable LO-type design potentially has a better receiver noisetemperature since a narrower IF bandwidth is sufficient, andthus, a lower noise figure LNA may be used. Furthermore,the variable LO design should have a smaller sensitivity im-balance between the upper and lower sidebands. Nevertheless,characterizing the radiometer receiver passband is essential foraccurate retrieval, particularly for estimating LWP, due to theasymmetry of the absorption spectrum of liquid water around183 GHz. The passband of the four GVR receivers were mea-sured with a calibrated G-band synthesized signal source. Theresulting receiver frequency response relative to 183.31 GHzis shown in Fig. 3. The rapidly diminishing noise figure of theharmonic mixer above 14 GHz is evident in the skewed outsidepassbands.

2204 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 45, NO. 7, JULY 2007

Fig. 3. Frequency response of the four GVR double-sideband receiver chan-nels. The output power of the G-band source used to make these measurementsover the 163- to 203-GHz band was flat to within 0.5 dB. This measured fre-quency response is needed to optimize the LWP and PWV retrieval algorithms.

III. TEST RESULTS

A prototype ground-based version of the radiometer wascompleted in late 2004. In early 2005, the instrument wastested over a temperature range of −40 ◦C to +25 ◦C in anenvironmental chamber. The stability of the instrument wasalso characterized by measuring the brightness temperatureof a constant temperature (∼263 K) external absorber in thechamber. Rau et al. [5] recommend the use of Allan deviationfor measuring the stability of radiometers. Allan deviation(ADN ) for a radiometer is defined as the

√0.5 × rms differ-

ence between nonoverlapping adjacent N -point averages of theerror series associated with the radiometer measured brightnesstemperatures t(i), i = 1, . . . ,M , relative to the true (error free)T (i), such that

ADN =

√12

j=( M

N )−1∑i=1

[EN (Ni + i − N) − T

]2MN − 1

1/2

(1)where EN (i) =

∑j=i+N−1j=1 ((t(j) − T (j))/N).

The measured AD∆t as a function of ∆t is shown in Fig. 4,where ∆t is the acquisition time of N brightness temperaturesamples. The AD curves indicate that the precision of theGVR brightness temperature measurements can be reduced byaveraging up to about 8 min. Due to the slow (0.1 Hz) measure-ment rate of the ground GVR, with the mechanically rotatedreflector mirror, this time interval corresponds to averaging only43 data points. The radiometer was configured such that eachof these data points was acquired by integrating the detectedsignal for about 0.3 s. The difference in the precision of thevarious channels is due to receiver bandwidth (500 MHz at1 GHz compared to 2 GHz at 14 GHz) and to differencesin the subharmonic mixer noise figure and IF port matching,which rapidly degrades above 12 GHz, as shown in Fig. 5. Theresulting instrument stability is more than sufficient in practice,however, since the temporal variability of the atmosphere isusually several Kelvin in less than a minute even in clear, calm,and nonconvective conditions.

The instrument precision suggested in Fig. 4 is only validwhen the sky temperature is close to the temperature of

Fig. 4. Allan deviation of the GVR measured in a temperature-stable chamberwhile observing a −10 ◦C absorber.

Fig. 5. IF port standing-wave ratio (SWR) of the Virginia Diodes Inc.183-GHz subharmonically pumped front-end mixer.

the GVR calibration loads. The temperature of the absorberused to measure the Allan deviation of Fig. 4 is about−10 ◦C (∼263 K), which is not much colder than the 293 KGVR warm load. Measurement errors increase with coldertemperatures, such that a 50-K sky measurement is expectedto have close to six times the errors shown in Fig. 4.

In February 2005, after a series of tests, the ground GVR wasinstalled on the roof of ProSensing facility, Amherst, MA, andleft to collect data continuously for three weeks. The antennawas kept clear of snow and debris using a combination ofa tilted 1-mil Mylar film window and a 500-ft3 · min (cfm)blower and hood. In mid-April 2005, the instrument was de-ployed at the DOE ARM program’s North Slope of the Alaskasite, as shown in Fig. 6, where it collected data continuously forone year.

Key system parameters for the GVR instrument are summa-rized in Table I.

IV. EXAMPLE DATA

On February 18, 2005, the GVR was operating from the roofof the ProSensing facility. In early afternoon, broken clouds, as

PAZMANY: COMPACT 183-GHz RADIOMETER FOR WATER VAPOR AND LIQUID WATER SENSING 2205

Fig. 6. Prototype ground-based GVR at the DOE North Slope of Alaska“Great White” site near Barrow, AK. A tilted 1-mil Mylar film radome windowis kept clear from rain or snow with a 500-cfm blower and hood.

TABLE IGVR KEY PARAMETERS

shown in Fig. 7, containing supercooled liquid passed above thezenith-pointed radiometer. The surface temperature was about+5 ◦C, and the clouds were approximately 1–2 km above theinstrument. The corresponding hour-long brightness tempera-ture data from the four channels are shown in Fig. 8. The PWVand LWP, as shown in Fig. 9, were estimated with two separateneural networks using the four brightness temperatures and thesurface air temperature as inputs. The neural networks usedfor this retrieval were trained with PWV and LWP computedfrom an atmospheric model generated using simulated liquidclouds combined with radiosonde data that were collected overa seven-year period in Albany, NY. For each computed PWVand LWP, the corresponding brightness temperature trainingdata at the four radiometer channels were calculated using theatmospheric absorption models compiled by Ulaby et al. [6].Validation of the data collected with the GVR in Barrow, AK,using coincident radiosonde data and the Rosenkranz correctedvapor absorption model [7] is presented by Cadeddu et al. [8].

The data presented here are a qualitative example of GVRsensitivity to PWV and LWP. Atmospheric conditions werequite favorable for precise PWV and LWP retrieval since the at-mosphere was sufficiently dry; thus, none of the channels weresaturated. The resulting precision of the retrieved LWP was lessthan 0.005 mm, and the precision of the estimated PWV wasabout 0.1 mm. Cadeddu et al. [8] have investigated the absolutecalibration of the GVR data. The said study concluded that the

Fig. 7. Wintertime fair-weather cumulus clouds passed over the GVR onFebruary 18, 2005, in Amherst, MA. Based on a +5 ◦C surface temperature andan estimated cloud altitude of 1–2 km, it is assumed that these clouds containedsupercooled liquid. The retrieved LWP data, as shown in Fig. 9, demonstratethe sensitivity of the instrument to the passing clouds.

Fig. 8. Data collected with the GVR from clouds containing supercooledliquid water, as shown in Fig. 7.

brightness temperatures measured during the dry winter monthsare in good agreement (within a few Kelvin) with brightnesstemperatures calculated based on radiosonde data.

V. DISCUSSION

The potential high sensitivity of radiometers operating nearthe 183-GHz vapor line to PWV and LWP has been convinc-ingly demonstrated by previous instruments and field cam-paigns [2]–[4]. The novelty of the radiometers described in thispaper is that it realizes this potential in a simple and compactdesign.

The method of using external calibration loads to trackreceiver gain and offset drifts appears to be very accurate evenat a low (∼0.1 Hz) calibration rate. The convection-heatedenclosure is quite large compared to the rest of the instrument,but it is necessary for absolutely calibrating the measurements.

Interference caused by a nearby high-power radar was foundto be a problem while operating at the Barrow, AK site. Metal

2206 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 45, NO. 7, JULY 2007

Fig. 9. Retrieved PWV and LWP of the fair-weather cumulus clouds shownin Fig. 7. Data products were estimated using a neural network algorithm fromthe measured brightness temperatures of Fig. 8 and surface temperature.

shielding around the IF section reduced the interference butdid not eliminate it. Fortunately, the interference was onlystrong enough to be noticeable when the scanned radar beamwas pointed directly at the radiometer, which only occurredonce every 2–3 min and effected only one data point perradar scan. Consequently, those corrupted data points could bereliably detected and eliminated using the conservative smooth-ing algorithm [9]. No other interference problems have beenencountered to date with GVR.

The combination of high-performance blower and Y-shapedpipe hood solved the problem of keeping the Mylar win-dow clear from ground debris and precipitation. Conventionalradiometer fan designs, which blow horizontally over theradome, would have ripped the 1-mil thin Mylar film in a fewdays, whereas, in spite of the more than 10 m/s updraft in theY-shaped pipe, the Mylar film at the base stayed intact after afull year of continuous operation. The 15-cm (6-in)-diameterY-shaped pipe also did not have a detectable effect on themeasured data, which is likely due to the narrow beam andhigh beam efficiency of the radiometer antenna. Nevertheless,the final system calibration, using external high-precision hotand warm loads, was conducted with the Y-shaped pipe hoodin place.

ACKNOWLEDGMENT

Many individuals have contributed to this instrument devel-opment effort. The author would like to thank A. Tanner forhis technical guidance during the design phase; J. Mead for

technical discussions throughout the project and for editingthe manuscript; J. Liljegren for organizing the deployment ofthe GVR to the ARM site, Barrow, AK; and W. Brower formaintaining and helping to calibrate the instrument in Alaska.The author would also like to thank R. Cochran, T. Chambers,and E. Black for developing the software; M. Cunningham,R. Lamoreux, and F. Leaf for constructing the instrument;G. Lee for the mechanical design, and B. Volain for designingand programming the FPGA data acquisition board.

REFERENCES

[1] P. E. Racette, E. R. Westwater, Y. Han, A. J. Gasiewski, M. Klein,D. Cimini, D. C. Jones, W. Manning, E. J. Kim, J. R. Wang, V. Leuski,and P. Kiedron, “Measurement of low amounts of precipitable water vaporusing ground-based millimeterwave radiometry,” J. Atmos. Ocean. Tech-nol., vol. 22, no. 4, pp. 317–337, 2005.

[2] A. Tanner, “Development of a high-stability radiometer,” Radio Sci.,vol. 33, no. 2, pp. 449–462, 1998.

[3] M. C. Wiedner, “Atmospheric water vapour and astronomical millimeterinterferometry,” Ph.D. dissertation, Univ. Cambridge, Cambridge, U.K.,1998.

[4] E. R. Westwater, D. Cimini, V. Mattioli, A. Gasiewski, M. Klein,V. Leuski, and J. Liljegren, “The 2004 north slope of Alaska arctic winterexperiment: Overview and highlights,” in Proc. MicroRad Spec. Meeting,2006, pp. 77–81.

[5] G. Rau, R. Schieder, and B. Vowinkel, “Characterization and measurementof radiometer stability,” in Proc. 14th Eur. Microw. Conf., Liege, Belgium,Sep. 10–13, 1984, pp. 248–253.

[6] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing,vol. I. Reading, MA: Addison-Wesley, 1981.

[7] P. Rosenkranz, “Water vapor continuum absorption: A comparison of mea-surements and models,” Radio Sci., vol. 33, no. 4, pp. 919–928, 1998.

[8] M. P. Cadeddu, J. C. Liljegren, and A. L. Pazmany, “Measurements andretrievals from a new 183-GHz water vapor radiometer in the Arctic,” inProc. MicroRad, 2006, pp. 252–255.

[9] A. Jain, Fundamentals of Digital Image Processing. Englewood Cliffs,NJ: Prentice-Hall, 1986, ch. 7.

Andrew L. Pazmany (S’90–M’93) received theB.S., M.S., and Ph.D. degrees in electrical engineer-ing from the University of Massachusetts, Amherst,in 1986, 1988, and 1993, respectively.

From 1993 to 2004, he was first a PostdoctoralResearcher and then a Research Associate Professorwith the Microwave Remote Sensing Laboratory,University of Massachusetts. He is currently a SeniorEngineer with ProSensing Inc., Amherst, developingcustom radar and radiometer systems for environ-mental remote sensing applications. His research pri-

marily focused on developing remote sensing and signal processing techniquesfor atmospheric remote sensing.