Comparison of columnar water vapor measurements during the fall 1997 ARM Intensive Observation Period: solar transmittance methods. B. Schmid, _J. J. Michalsky, 2 D. W. Slater, 3J. C. Barnard, 3R. N. Halthore, 4J. C. Liljegren: B. N. Holben, 6T. F. Eck, 7 J. M. Livingston, s P. B. Russell, 9T. Ingold 1°, and I. Slutsker 1_ tBay Area Environmental Research Institute, 3430 Noriega Street, San Francisco, CA 94122. (e-mail: bschmid @mail.arc.nasa.gov) 2Atmospheric Sciences Research Center, State University of New York, Albany, 251 Fuller Road, Albany, NY 12203. (e-mail: [email protected]) 3pacific Northwest National Laboratory, P.O. Box 999, Richland, WA 99352 (e-mail: donald.slater @pnl.gov, james.barnard @pnl.gov) 4Brookhaven National Laboratory, P.O. Box 5000, Upton, NY 11973. (e-mail: [email protected]) 5Ames Laboratory, Ames, Iowa, USA Now at: Argonne National Laboratory, 9700 South Cass Ave., Argonne, IL 60439 (e-mail: [email protected]) 6NASA Goddard Space Flight Center, Code 923, Greenbelt, MD 20771, (e-mail: brent @aeronet.gsfc.nasa.gov) 7Raytheon ITSS/NASA Goddard Space Flight Center, Code 923, Greenbelt, MD 20771, (e-mail: [email protected]) sSR/International, 333 Ravenswood Avenue, Menlo Park, CA 94025. (e-mail: jlivingston @mail. arc. nasa. gov) 9NASA Ames Research Center, MS 245-5, Moffett Field, CA 94035-1000. (e-mail: prussell @mail. arc. nasa. gov) t°Institute of Applied Physics, University of Bern, Sidlerstr. 5, CH-3012 Bern, Switzerland. (e-mail: ingold @mw.iap.unibe.ch) _ISSAI/NASA Goddard Space Flight Center, Code 923, Greenbelt, MD 20771, (e-mail: [email protected]) Submitted to Applied Optics on June 9, 2000 Abstract. In the fall of 1997, during an Intensive Observation Period (IOP), the Atmospheric Radiation Measurement (ARM) program conducted a study of water vapor abundance measurement at its Southern Great Plains (SGP) site. Among a large number of instruments, four sun-tracking radiometers were present to measure the columnar water vapor (CWV). All four solar radiometers retrieve CWV by measuring total solar transmittance in the 0.94-_tm water vapor absorption band and subtracting contributions due to Rayleigh, ozone and aerosol transmittances. The aerosol optical depth comparisons among the same four radiometers has been presented elsewhere (Geophys. Res. Lett., 26, 17, 2725-2728, 1999). We have used three different methods to retrieve CWV. In a fu'st round of comparison no attempt was made to standardize on the same radiative transfer model and its underlying water vapor spectroscopy. In the second round of comparison we used the same line-by-line code (which includes recently corrected H20 spectroscopy) to retrieve CWV from all four sun- tracking radiometers. This decreased the mean CWV by 8% or 13%. The spread of 8% in the solar radiometer results found when using the same model is an indication of the other-than- model uncertainties involved in determining CWV from solar transmittance measurements with current instrumentation. OCIS codes: 010.0010, 010.1110, 010.1320, 010.7340 1. Introduction Solar transmittance methods can provide water vapor abundance from direct or reflected sunlight measurements in spectral channels in and adjacent to water vapor absorption bands. The so-derived water-vapor transmittance has to be translated into columnar water vapor (CWV). Although this relationship is well known qualitatively, I it has proven difficult to quantify. Attempts to do so for water-vapor absorption bands in the near-infrared date back to 1912. 2But 2 evenin the last decadetherehasbeena steadystreamof publicationson this subject.For example,resultsfromground-basedretrievalsof CWV usingsunphotometer(sSPM) havebeen reportedwidely (seeIngoldet al. 3 and references therein). Recently, Schmid et al.4 reported on CWV retrievals using an airborne sunphotometer. Instruments aboard satellites, such as SAGE II (Stratospheric Aerosol and Gas Experiment) and POAM II and III (Polar Ozone and Aerosol Measurements) use the solar occultation technique (i.e., they act like a SPM by measuring the solar transmittance through the limb of the atmosphere) to retrieve water vapor. 5"6Finally, CWV is also retrieved from airborne (such us AVIRIS (Airborne Visible Infra Red Imaging Spectrometer)) and spaceborne (such as POLDER (POLarization and Directionality of the Earth's Reflectance) or MODIS (MODerate-resolution Imaging Spectroradiorneter)) instruments that measure the solar radiance reflected by the Earth surface. 7'8'9a° Recent findings that the H20 line intensities in the visible and near infrared portion of the widely used HITRAN-96 database 1_were in error _2and that H20 lines (especially weak ones) might be missing from the current databases 13'14have sparked renewed discussion of the accurate conversion of measured water-vapor transmittance into amounts of water vapor. In the fall of 1997 the Atmospheric Radiation Measurement (ARM) program _5conducted the 24 Intensive Observation Period (lOP) to study water vapor at its Southern Great Plains (SGP) site. Among a large number of systems such as radiosondes, microwave radiometers, raman lidars, Global Positioning System receivers, and an infrared spectrometer, four sun-tracking radiometers were present to measure water vapor. ,6 In this paper we focus on the four sun-tracking radiometers that retrieve CWV by measuring solar transmittance in the 0.94-I.tm water vapor absorption band. The measurements were made between 15 September and 5 October 1997 at the SGP ARM central facility near Lamont, Oklahoma(36° 36' N, 97° 22' W, 316 m above sea level). Dry to very humid conditions, with CWV ranging from 1to 5 cm, were experienced over the three-week period. As one of the steps in the CWV retrievals the aerosol component must be subtracted from the total transmittance in the 0.94-ktm band. The aerosol optical depth (AOD) comparison among the same four radiometers has been presented previously, t7 Following the philosophy of the just-mentioned AOD-comparison we fh'st made no attempt to standardize on the methods used to derive CWV from the four radiometers. We found that three different methods had been used in conjunction with three different radiative transfer models. In a second round we used the same radiative transfer model (with its underlying spectroscopy corrected according to Giver et al. 12) for all instruments. In this paper we will show the results from both comparisons. 2. Instrumentation The NASA Ames Research Center deployed its six-channel Ames Airborne Tracking Sunphotometer (AATS-6) at the SGP central facility of ARM for this IOP. This instrument, described by Matsumoto et al., 18uses an active sun sensor to keep the instrument pointed at the solar disk. The central wavelengths and full widths at half maximum (FWHM) for the filters are given in Table 1. The Si detectors are held at a constant temperature of 45:t: 0.6 °C. The field-of- view (FOV) of AATS-6 is 4.5 °. A measurement sequence was repeated every 12 seconds with all filters scanned nine times then averaged in the first three seconds of the 12-second period. At the ARM SGP central facility a CIMEL sun/sky photometer measures AOD. This instrument is also part of AERONET, a worldwide network of CIMEL sunphotometers [Holben et al., 1998]. _9The CIMEL CE-318 points to the sun based on an ephemeris calculation and then fine tunes the pointing with an active sun sensor adjustment. Samples consist of triplets of measurementswith eachmemberof the triplet beginning30 secondsapartandconsistingof eight filter measurementscompletedwithin eight seconds;the triplets are repeatedat every quarterairmassbetweentwo tosevenairmassesandevery15minuteswhentheair massis less thantwo.ThecentralwavelengthandFWHM foreachfilter aregivenin Table1.Thefield-of- viewis 1.2°.Thetemperatureoftheinstrumentismonitoredbutnotcontrolled. The multi-filter rotatingshadowbandradiometer(MFRSR)2°hasa hemisphericalfield-of- view. A bandis positionedto alternatelymovecompletelyoutof the field-of-viewandthento blockthe sunaccordingto a solarhouranglecalculationallowing a measuremenotf the total downwardanddiffusedownwardirradiance.Thedifferencebetweenthe two measurementsis thedirectsolarcomponentnormaltothereceiver,andthedirectnormalcomponentiscalculated bydividing bythecosineofthesolar-zenithangleandcorrectingfortheangularresponseof the quasi-Lambertiandetector.Samplingis every20seconds.ThecentralwavelengthandFWHM foreachfilter aregiveninTable1.Thetemperatureisheldat40°C. The rotating shadowbandspectroradiometer(RSS)21has a Lamhertianreceiver and a shadowingsequencesimilarto theMFRSR;however,thedetectoris a512-elementphotodiode array that receivesits energy input from the focus of a prism spectrograph.Samplingis performedonceeachminute.Thespectralresolutionbetween350and1050nmdiminishesfrom 0.3to8nmbecauseof theprismdispersiveelement.Thetemperatureisheldat40°C. Inthefollowing wewill refertoallfourinstrumentsassunphotometer(sSPM). 3. Methodology In the derivation of atmospheric transmittance, we distinguish between atmospheric window channels and gaseous-absorption channels. The window channels are located outside of molecularabsorptionbandssuchasO2or H:O bands,andarenormallyusedto determinethe aerosolopticaldepth. 3.1.AerosolOptical Depth ForatmosphericwindowchannelstheSPMoutputvoltage,V(;t,), obtained when observing the directly transmitted solar irradiance over a small bandpass AA centered at wavelength A can be described by the Beer-Lambert-Bouguer attenuation law vo(;t)R-2 (1) where Vo (;t) is the instrurr_nt calibration constant, R is the Earth-Sun distance in astronomical units (AU) at the time of observation, _'(;t) is the spectral optical depth, and m is the relative optical airmass, a function of the solar zenith angle. Taking the logarithm of (l) leads to lnV(A) = In[Vo(A)R-2]- mr(A) (2) If a series of measurements is taken over a range of airmasses m during which the optical depth _'(A) remained constant, Vo (A) may be determined from the ordinate intercept of a least-squares fit when plotting the left-hand side of (2) versus m. This procedure is commonly known as Langley-plot calibration. In (1) several attenuators contribute to v(2): "r(_.) = ZR(X) +'r 3(,,1,)+ Zz(;t) + z,, (A,) (3) where the subscripts R, 3, 2 and a refer to Rayleigh scattering by air molecules, absorption due to 03 and 02, and attenuation due to aerosol particles, respectively. A refined Langley technique z2a3"24 - which uses individual airmass expressions for each attenuator in (3) - was used for AATS-6 but not for the other 3 instruments. The window channels of AATS-6 were calibrated by averaging the results of 6 successful morning Langley 6 plotsperformedattheMaunaLoaObservatory(MLO) inHawaii(19° 32' N, 155° 34' W, 3397 mabovesealevel)abouttwo weeksbeforethelOP. Calibrationof Cimel #27(the instrumentdeployedat SGPduringthe IOP) is basedon a transferof thecalibrationfromCimel#37,thereferenceinstrument.The intercalibrationswere performedat GoddardSpaceFlight Centerin Marylandon 30 August 1997and3 November 1997atmiddayforaperiodof 1-2hours.Thereferenceinstrumentitselfwascalibratedusingthe LangleytechniqueatMLO inMayandSeptember1997. Calibration of MFRSR andRSS was basedon a robustestimateusing the 20 nearest successfuLl angleyplotsatSGP.Oneof those20nearestsuccessfuLlangleyplotswasobtained with datafrom the morningof 29 September1997.A Langleyplot performedwith AATS-6 duringthatsamemorningyieldedcalibrationconstantsthatagreedwithin 0.5%with theMauna Loaresultsobtainedtwo weeksbeforethelOP.Thissuggeststhatduringthis particularmorning theatmosphereoverSGPwassufficientlystabletoyieldunbiasedLangleyplotresultstobeused intherobustestimateofthecalibrationconstantsforMFRSRandRSS. OncethecalibrationconstantsVo (2) of the window channels are known the aerosol optical depth _'a('_') call be determined from (2) - (3). The AODs obtained from each instrument were derived independently of one another. Although the methods to remove Rayleigh, ozone and nitrogen dioxide optical depths may coincide in some instances, there was no attempt at a uniform reduction to aerosol optical depth from total optical depth. Nevertheless, AODs 0,,=380- 1020 nm) obtained during the lOP by Cimel, MFRSR and RSS agreed with AATS-6 values to within 0.025 (rms). The AODs in atmospheric "'windows" adjacent to the 0.94-1am band agreed within 0.015 (rms). 17 3.2.Columnar WaterVapor The Beer-Lambert-Bouguer law, monochromatic in its nature, may be applied over small bandpasses aX with negligible error as long as the spectral variation of transmittance inside the bandpass is small. In regions of strong spectral variation of molecular absorption, such as the near-infrared water-vapor absorption bands, (1) may be expressed as25 v(z)-- (;t)+ (X) (4) (Note that there is no absorption due to NO2 in the water-vapor absorption channels used here). Tw(_-) is the band- and source-weighted water-vapor transmittance Tw(A-)- f,_xE°(_')S(A)exp[-m'rw(_)]d'_ (5) _azEo (3,)S(A)dA where rw(A) is the strongly varying water vapor absorption optical depth, E0(A,) is the exoatmospheric solar irradiance, and S(_) is the instrument response. It should be noted that even if E 0(2) and S(_,) were effectively constant over A2, the strong spectral variation of • w(_,) is sufficient to require the band-weighted transmittance Tw(_-) in (4). Also, (4) does not follow the Beer-Lambert-Bouguer law, as Tw(_ ) generally cannot be modeled by an exponential with a negative argument of airmass times a constant band-weighted optical depth. Hence, for channels in strong absorption bands, Vo (A) can no longer be found using the traditional or refmed Langley method. In this paper we discuss three different approaches to determine Vo (2) and Tw(X) in order to determine CWV from measurements in the 0.94-1am water vapor absorption band. Method A: Modified Langley plot technique If Tw(_-) can be modeled by an exponential with a negative argument proportional to some power of the slant path absorber amount such as Tw(_)=exp_a(muf] (6) where u is the columnar water vapor and a and b are constants, then Vo ().) can be determined using a modified Langley plot technique: Substituting (6) into (4), rearranging the terms and taking the logarithm leads to In V(_)+ m['Ca(,_)+'CR(_,)+'C3(_,)]=ln_o(_)R -2 ]-a(mu_ (7) Modified Langley plots are now constructed by plotting the left-hand side of (7) versus mb. Therefore, the instantaneous values of the aerosol optical depth ra(A) in the water-vapor absorption channels are needed. These are estimated from the SPM "window" wavelengths using a quadratic fit on a log-log scale of r_(2) versus ;_. This requires the Vo(2) values of the "window" channels to be determined before constructing modified Langley-plots. It is evident that for the construction of modified Langley plots the columnar water-vapor amount should remain constant, at least for the 1.5 to 2 hour period of Langley data acquisition. Tw(_') is typically computed according to (5) over a range of slant path water vapor amounts using a radiative transfer model. The constants a and b in (6) are then found by a curve-fitting procedure. 3'26'27"2sCombining equations (4) and (6) the CWV is 1 U=mL-a_'V--_l'l_l(inV°(_-'2)R m[,rR(_,)+,t.a(,_)+,t.3(X)])] _ (8) In this paper we used method A to obtain CWV for AATS-6 and Cimel. For the Cimel instrument the standard AERONET algorithm was used: the same typical filter function (S(&) in (5)) was used for all instruments in the network in conjunction with LOWTRAN7 computations 29to determine one set of a and b. The 940-rim channel of the reference instrument Cimel #37 was calibrated using the modified Langley technique at MLO in May and September 1997. For both calibration periods the Vo (A,) values of 4 morning modified Langley plots were averaged. The relative standard deviations in Vo (&) were -2%. For AATS-6 we used MODTRAN 3.5 vl.129 to determine one set of a and b values for MLO and several sets (covering different ranges of mwu ) for SGP conditions. For S(2) in (5) we used the filter function of the 941.4-nm channel as measured by the manufacturer (Barr Associates Inc., Westford, MA) in February 1994. The Vo (A) value of that channel was determined by averaging the results of 5 morning modified Langley plots (standard deviation 1.2%) performed at the Mauna Loa Observatory (MLO) two weeks before the lOP. Method B: Differential lamp/solar spectrum technique This method has been described in detail by Michalsky et al. 3° Only a brief summary is given here. Method B avoids the need to calibrate using the modified Langley method. Instead, it requires the instrument output VL(A) when viewing a calibration lamp, the lamp irradiance EL(_,) and the extraterrestrial solar spectrum Eo(2), both convolved with the filter function S(_). In order to retrieve CWV we consider the ratio of the SPM output voltages measured in channels in (;tin) and adjacent to (_,o,a) the 0.94-_tm band _ Eo(Z,n)e(L o )VL ))](9) I0