simultaneous remote measurements of atmospheric temperature and humidity using a continuously...

Simultaneous remote measurements of atmospherictemperature and humidity using a continuouslytunable IR lidar

M. Endemann and R. L. Byer

The possibility of accurate remote measurements of temperature and humidity using a three-wavelengthdifferential absorption technique is discussed. Selection of water vapor absorption lines for the measure-ment promises highest accuracy for temperature measurements and also allows the simultaneous determina-tion of absolute humidity. Preliminary measurements of average temperature and humidity over a 775-mlong path using a continuously tunable infrared lidar are reported. For the temperature measurements arelative accuracy of 1.40C was observed. The absolute error, however, is presently -5 0C due to inaccuraciesin the wavelength selection of the lidar transmitter. Humidity was measured with a 1.5% relative error lim-ited by return signal fluctuations. The method is feasible for depth resolved measurements of temperatureand humidity.

1. Introduction

Interest in remote temperature measurements hasgrown in recent years. Measurement of atmospherictemperature profiles with a 2-km range resolution and1°C accuracy is required to allow better understandingof weather. Presently passive temperature soundingtechniques allow profiling with a resolution of 1-2 scaleheights (8-16 km) and a 2.50 C accuracy.

Several temperature probing techniques using lightbeams have been proposed. As early as 1953 Eltermanmeasured stratospheric temperature profiles with asearchlight technique.1 2 Temperature was obtainedfrom measured density variations. With searchlights,this procedure is not applicable in the troposphere dueto unknown aerosol particle scattering. Laser tech-niques have been proposed using Raman scattering 3 orthe use of different wavelengths4 to determine atmo-spheric density fluctuations and thus derive tempera-ture profiles. However, the calibration of the obtainedtemperature data must be done with an independentmeasurement.

When this work was done both authors were with Stanford Uni-versity, Applied Physics Department, Ginzton Laboratory of Physics,Stanford, California 94305; M. Endemann is now with Battelle Inst-itut e.V., 6000 Frankfurt a.M. 90, Federal Republic of Germany.

Received 2 February 1981.0003-6935/81/183211-07$00.50/0.© 1981 Optical Society of America.

Calibrated temperature measurements are possiblewith a technique proposed by Cooney.5-7 The rota-tional Raman spectrum of nitrogen allows the deter-mination of atmospheric temperatures with good ac-curacy. Recent experiments8 demonstrated depth re-solved temperature measurements to 2300-m altitudewith a 1C relative error. Although in the reportedmeasurements the data were calibrated with measure-ments from a radiosonde, absolute temperature mea-surements are possible with this technique.

The temperature dependence of the population inmolecular levels can be used to probe atmospherictemperatures with absorption measurements.9"10 Forremote temperature measurements using a lidar threedifferent wavelengths are required. Two frequenciesare transmitted on the absorption lines of a species,while the third one is used for calibration of the lidar.A computer simulation of this method shows thattemperature profiling from the ground or an airplaneis feasible."

Only a few experiments have been reported to sup-port the theoretical analyses. Murray et al.12 used aCO2 lidar to measure the average temperature over a5-km long path. However, only two frequencies wereused in this measurement, and the temperature wasderived under the assumption that the partial pressureof CO2 in the atmosphere is constant. Correlation ofthe remotely measured temperature with thermometermeasurements shows the feasibility of the method, al-though absolute errors in excess of 5C occurred.Kalshoven et al. 13 also reported a two-frequency mea-surement of temperature using a cw dye laser to probe

15 September 1981 / Vol. 20, No. 18 / APPLIED OPTICS 3211

oxygen absorption lines.14 A relative error of 0.50C wasobserved in measurements over a 1-km long path.

We report single-ended remote temperature mea-surements using a three-frequency method. A theoryis derived which allows the use of species with complexabsorption spectra and assessment of temperature ac-curacy. We selected H20 as the absorbing species forour measurements. Water vapor not only promises thehighest sensitivity for a remote temperature measure-ment, but it also allows the simultaneous measurementof absolute humidity which is of interest for meteo-rology.' 5

II. Theory

The temperature dependence of the population inmolecular levels causes a variation of an absorptioncross section o(v,T) at frequency v with temperature T.The absorption cross section is given by

a(v,T) = S(v,T)/ry(T), (1)

where y(T) is the halfwidth at half-maximum of theabsorption line. The temperature dependence of theabsorption line strength S is given by the Boltzmanndistribution exp(-E 0 /kT) and a partition functionQ(T) by16

S(vT) = S(vTo) Q(To) exp E 1 1)] (2)Q(T) k T T

for an arbitrary reference temperature To, where E isthe lower state energy of the transition at frequency vi,and k is the Boltzmann constant. The temperaturedependence of y(T) can be described by

'y(T) = y(To) (To/T)n. (3)

In the Doppler broadened regime, n is independentof the transition and given as 1/2. If the collision pa-rameters are temperature independent n is 1/2 also in thepressure broadened regime. In general, however, theparameter n depends on the transition. Benedict andKaplan17 numerically calculated values of n for watervapor pressure broadened with N2. They found valuesranging from 0.756 to -0.045 dependent on the line-width of the transition.

The temperature dependence of the absorption crosssection is then

a(v,T) = o(vTo) (-T)-n Q(T) -exp [ k (o (4)FT TI Q(To) 1k T I

In atmospheric measurements the absorption crosssection is not determined directly. The differentialabsorption measurement determines the absorbanceA(v,T) 2N R- o-(v,T) due to the absorption line.Here N is the number density of the absorbing species,and R is the range over which the measurement ismade.

The absorbance can be measured using the lidarprinciple. Two different measurement modes arepossible18: long path average measurements of theabsorbance using topographic targets as the noncoop-erative backscatterer or depth resolved determinationof the absorbance using Rayleigh and Mie scattering asthe reflector. Vertical profiling and remote mapping

of temperature are possible with the latter mode ofoperation. In the visible and UV, depth resolvedmeasurements over a 1-km long path require trans-mitted energies of the order of 5 mJ. In the infrared,however, decreased backscattering coefficients anddetector sensitivity require about 100 times that energyfor the same performance. The SNR of the receivedsignal can be improved by averaging over several laserpulses.

To remotely determine the temperature the factorN-R must be measured independently. This can bedone using another absorption line of the same species.The measurement of temperature then requires thetransmission of three different frequencies, v1 and v2 onthe selected absorption lines, and 3 as reference tocalibrate the return signals. From the received signalsR (vi) at the different frequencies we determine theabsorbance as A(vi,T) = 1n[R(v)/R(v 3 )] for i = 1, 2.Then the temperature is given as

T= To {1 k~'1AE. [n .I) + T / ln 2 ( (5)

where Ai = A(vi,T), o-i = or(vjTo), AEO = E- E2 is thedifference in lower level energies, and An = ni - n 2 isthe difference of the exponents describing the temper-ature dependence of the linewidths. In the derivationof Eq. (5) we used the linear approximation for thelogarithm of T/To. If we further linearize Eq. (5) weobtain, for Al/A 2 << 1/0f2 exp(AEO/kT) or T - To <<TO,

T= C ln (2) + D,

withkT.2. kT 0 - (a2l

CAE (1 + An.) , D= C ln n + To. (6)

The constant D can be found either using spectroscopicdata, or it can be determined empirically with calibratedtemperature measurements.

It is straightforward to calculate the density of thespecies from the measured absorbances if the absolutesize of the absorption cross sections is known. Thismeans that the temperature measurement and theconcentration of the absorbing species can be probedsimultaneously.

The accuracy of a remote temperature measurementis determined by the error of each absorbance mea-surement. If we express the uncertainty in each mea-surement by the SNR of the least attenuated signal, wefind the temperature uncertainty AT for small An as

T = ko. 1 [2 1 + exp2Al 1/2s\E S/ =1 A?2 }

(7)

We see that the accuracy of the measurement is in-versely proportional to the difference in lower levelenergies. It can also be noted that the relative errorATIT - AT/To increases linearly with reference tem-perature To.

3212 APPLIED OPTICS / Vol. 20, No. 18 / 15 September 1981

N

+

0.

0

0Ir

0-

sr

0 0.5 1.0 1.5 2.0 2.5ABSORBANCE A -

3.0

Fig. 1. Dependence of temperature accuracy on absorbance.

In addition the temperature accuracy depends onabsorbance due to the probed lines. As can be seerFig. 1 this error factor rises rapidly for absorbance,< 0.5 and A > 2. The best temperature accurac3achieved for A - 1.1. Then Eq. (7) simplifies to

AT cx4 . ° .2AEO S/N

We see that it is important to select lines with enotstrength to allow an appropriate absorbance over alected path. A tunable laser source is in general nessary to select such absorption lines.

The maximum lower level energy differences inbrational-rotational bands are typically -800 cmFor probing atmospheric temperatures where To K, the proportionality constant kT2/AE 75°C.the best conditions, we need a SNR of 300 for a teperature accuracy of 1C. An additional difficultythe use of a rotational absorption band is the 30-cnseparation of the lines with large lower level eneidifferences. This makes tuning between them a di,cult task. Generally two laser systems must be usecthis case.

Some of these difficulties can be overcome by usiwater vapor as the absorbing species. Since the atnspheric water vapor concentration is in the vicinity1%, lines with high lying lower energy levels can still lto rather strong absorption over atmospheric patWe can also expect to locate lines with higher lower leenergy states than with other atmospheric constitueand so achieve a lower required SNR for the measument. In addition, many vibrational-rotational baroverlap in the spectrum of water vapor which make!possible to select lines with large lower level eneidifferences that are only 1 cm-' apart. Furthermothe selection of water vapor allows the simultane(measurement of temperature and absolute humidiand so the determination of relative humidity is psible.

In addition to temperature errors due to the SNRthe received signals other noise sources are importanWe will discuss those in more detail in the next sition.

Il. Measurements

For our measurements we used a continuously tun-able infrared lidar. Figure 2 shows a schematic of thislidar. It has been described in more detail in anotherpaper. 2 0

An optical parametric oscillator (OPO) is used as thetransmitter source. This device is unique in having a4500-cm-1 wide tuning range from 1.4 to 4.0 im. Anunstable resonator Nd:YAG laser is used as a pumpsource. The doughnut shaped output profile of theunstable resonator beam is transformed in a multiplepath cell into a nearly Gaussian profile which allowsmore efficient and more reliable operation of the OPO.An angle phase matched LiNbO 3 crystal is the gain el-ement of the OPO. Output energy is 5 mJ in a 10-nsec

A- long pulse. The pulse repetition rate is 10 Hz.The output spectrum is controlled by an intracavity

grating with 600 grooves/mm and a 2-mm thick tiltedthe etalon. The linewidth of the OPO output is 0.08 cm-'.in Tuning of the lidar transmitter is done under computerA control. The LiNbO3 crystal, grating, and tilted etalon

y is are adjusted by a PDP 11/10 minicomputer in an openloop control system. The computer is also used to ratio,average, process, and display data taken with the lidar.

(8) After alignment of the transmitter, the computer con-trol allows automatic measurements.

igh The output beam from the OPO is collimated andse- expanded to a 2.5-cm diam beam and transmittedec- coaxially from the receiving telescope on the roof of the

laboratory. This telescope is a 40-cm diam Newtonianvi- design. It focuses the received light onto a liquid ni--'. trogen cooled InSb detector.290 A reference detector in the laboratory is used toIn monitor pulse-to-pulse fluctuations of the output en-rm- ergy. Its output signal is used by the computer to ratioin the received signal and so obtain energy independent1'- measurements.Tgy The 5-mJ output energy of the transmitter does notiffi- allow depth resolved measurements. We used ain building located 775 m from the telescope as the non-

cooperative target.ing We used tables of water vapor absorption lines of thelo- 1.9-gm2l and the 2.7-/,m22 bands in addition to theof AFCRL tapes' 6 for the selection of appropriate lines.

,achs.velts

re-idsit

rgyre,)usty,os -

,of.19

ec-

InSb DETECTOR [Ii40 cm DlAM. _ -RECEIVERTELESCOPE - ---

FROM OPOTO TELESCOPE

OPO

Fig. 2. Schematic of the continuously tunable infrared lidar.


-, E= 43cm 1 Eo=543cm-1

Eo=1999cm-1 E Xi999cm1

0

wzM0U.)c

5647 5650.41 5656 5647 5650A1 56565651.33 5651.33

FREQUENCY (cm 1 )

Fig. 3. High resolution photoacoustic spectrum of H2 0 around 5650cm-1 at different temperatures. The selected absorption lines aremarked. The upper trace shows a simulation using the AFCRL

tapes.1 6

For our purposes the ideal lines were found in the wingsof the 1.9-gm band. We selected aJ = 6 line of the vlv2combination band at 5650.41 cm-' and a J = 11 line ofthe 2V3 band at 5651.33 cm-' for our measurements.These lines are within 1 cm-' of each other and so alloweasy tuning of the OPO. Their lower level energies are543 and 1999 cm-'. The 1456-cm'1 energy differenceallows sensitive temperature measurements. In addi-tion, the line strengths at room temperature are -2.4and 1.1 X 10-3 (atm cm)-'. This allows measurementsover path lengths of 1 km with absorbances between 0.3and 1.0. For these conditions we calculate from Eq. (7)that a SNR of 200 is necessary for a temperature accu-racy of 1 0C. If the SNR for a single return signal is 25,we expect to average over sixty-four laser pulses for therequired accuracy.

We used the calculations from Benedict and Kaplan' 7

to obtain the exponents describing the temperaturedependence of the linewidth of the selected lines ac-cording to Eq. (3). They are 0.45 and 0.30, respectively.The correction factor An * (kTO/AE 0) in Eq. (6) is then2%. We can neglect this correction for the constants Cand D.

The first step in the measurement was the identifi-cation of the absorption lines in the laboratory and theverification of their temperature dependence. Thesmall absorption cross sections of the selected linesmade it difficult to measure absorption spectra in ashort path length. Since we were not primarily inter-ested in absolute absorption cross sections, we used aphotoacoustic cell for this work. The lower traces inFig. 3 show the signal from the photoacoustic cell filledwith water vapor and air for pressure broadening atdifferent temperatures. The upper traces show simu-lations of the spectrum using the spectroscopic data ofthe AFCRL tapes.' 6 Although the noise of the pho-toacoustic measurements is high, it can be seen that theselected absorption lines show the expected tempera-ture dependence. Also it can be noted that the ratio of

the line strengths at room temperature is different fromthat predicted using the spectroscopic data. Thischanges the calibration constant D given in Eq. (6) toa smaller value than predicted. In atmospheric mea-surements we used data from a recording thermographwith a +10 C accuracy to obtain D.

For the determination of the absolute humidity fromthe experimental data we used the absorption crosssection calculated from the AFCRL tapes' 6 for the5650.41-cm-1 line. The absolute cross-section accuracyof the AFCRL tapes limits the absolute accuracy of thehumidity measurement values to 20%.

It can be noticed in Fig. 3 that the two selected ab-sorption lines overlap. In addition more weaker lineswith different lower level energies contribute to theabsorbance at each of the wavelengths. For the refer-ence transmission measurement we selected a wave-length between the two temperature sensitive lines eventhough at this wavelength some light is absorbed bydifferent water vapor absorption lines. The absorptionof lines with different lower energy levels causes de-viations from the temperature inversion given by Eq.(5). However, we can expect to use the linear approx-imation given in Eq. (6) with slightly modified constantsC and D.

To find these new constants we used a computerprogram simulating the temperature measurements.The temperature dependence of the linewidths was notconsidered in this simulation. Figure 4 shows the cal-ibration curve obtained from the calculation. Here thetemperature dependence of the logarithm of the cal-culated absorbances T* is plotted for two differentlinewidths of the transmitted beam.

It can be seen that the curves deviate considerablyfrom the linear approximation given by Eq. (6) over the1000C temperature range. However, if a maximumtemperature error of 10C is allowed, the linear ap-proximation can be used from -10 to 30 0C. This rangeis wide enough to allow the use of Eq. (6) for the mea-surement of atmospheric temperatures in the lowertroposphere.

I

,z<

*F

-40 -20 0 20 40 60T (C)

Fig. 4. Simulation of the dependence of the logarithm of the ratioof the absorbances T* on the temperature.


VI 5651.33cm-1

'2: 5650.41cm-1"off 5651.OOcm-1

0

T CT*+ DC 36.8~D 48.4 C

RSOLUTION 0.1 cm-'

2 RESOLUTION 0.2 cm1

3

-

Z I~~II

[v5651.33 cm' [E0h 1999cm-

1

L, ~2 5650.41 cm1LE 2543 cm|

5650 5655 5660FREQUENCY (cml)

Fig. 5. High resolution atmospheric transmittance spectrum overa 775-m long path. The OPO linewidth is 0.1 cm-1 . The OPOmeasured spectrum in the lower trace is compared with a simulation

using the AFCRL tapes16 in the upper trace.

The constants found in the computer simulation forthe 0.1-cm-1 linewidth are C = 36.80 C and D = 48.40 C.Using Eq. (6) we obtain C = 40.40 C and D = 51.80 C.We notice an 8% decrease of the proportionality con-stant C due to the effects of other H20 absorption lines.This change improves the accuracy of the measurement.The change in D is unimportant since the line strengthsgiven in the spectroscopic data are not accurate enoughfor an absolute determination of temperature.

The line shape used in the simulation is differentfrom the actual line shape of the OPO output. To es-timate the error we repeated the simulation for differentlinewidths. We see in Fig. 4 that a 100% change inlinewidth leads to small changes in the linear approxi-mation. However, the proportionality constant Cchanges from 36.80C to 35.50C or 4%. This indicatesthat C is not strongly dependent on linewidth or lineshape of the OPO.

Two effects limit the accuracy of the OPO frequencyselection and thus affect the temperature measurement.The mechanical drive for the tilted etalon has a varyingstep size from 0.0002 to 0.03 cm-'. This means that thewavelength of the OPO can deviate from the intendedvalue by 0.015 cm-'. A second limitation on the tuningaccuracy is the axial mode spacing of the OPO. Themode spacing of the output at the resonated signal waveis 0.03 cm-. For operation with an extracavity tiltedetalon only three axial modes oscillate, and smallchanges of the cavity length change the frequency of thestrongest one. Both effects influence the accuracy ofa temperature measurement.

Drifting of the axial mode spectrum affects thewavelength of all three wavelengths for a temperaturemeasurement in the same way. We simulated this ef-fect and found that a frequency variation of 0.03 cm-'of all three wavelengths leaves the constants C and Deffectively unchanged. Therefore, we can neglect errorsfrom drifts of the axial mode spectrum of the OPO.

The change in step size of the etalon drive can causedeviations at only one of the measurement wavelengths.

While the variation of the reference wavelength betweenthe absorption lines shows only small effects on C andD, the change of any one of the two frequencies tunedon the absorption lines causes larger deviations. Theproportionality constant C changes 3% for a 0.03-cm-1variation of one of these frequencies. For measure-ments 200 C from To this results in a 1 0 C temperatureerror. However, this error can be neglected comparedto the change of the calibration constant D. A 0.03-cm-1 variation of one frequency alone causes deviationsin D as large as 5.20C. The mechanical design of theOPO transmitter allows frequency drifts of this mag-nitude, so deviations of the calibration constant overtime scales of tens of minutes can occur.

The lower trace in Fig. 5 shows an atmospherictransmission scan with a 0.1-cm-1 resolution tuned overa 17-cm-1 range around the selected lines. The uppertrace is a simulated transmission spectrum usingAFCRL tapes.' 6 We used scans similar to this to cali-brate the tuning of the OPO and to identify the tem-perature sensitive lines. Once the wavelength of theselines is determined and the OPO is aligned, measure-ments of temperature and humidity are done undercomputer control. The return signal at each frequencyis averaged over several shots before tuning to anotherwavelength. To keep errors due to water vapor fluc-tuations small, it is desirable to tune the OPO after eachshot and average the temperature data afterward.However, tuning between different wavelengths takesup to 0.5 sec. To improve the time resolution of ourmeasurements we average 30 shots per line beforetuning to the next line. When all three requiredwavelengths are probed, temperature and humidity arecalculated. This cycle continues until a total numberof 150 shots per line is reached. Then the computercalculates the average of five temperature and humiditydata points, as well as the error due to return signalfluctuations, and stores and displays those values.

For calibration of our temperature measurement weminimized the difference between the remote temper-ature data and data from a recording thermograph atthe telescope. Figure 6 shows the 8.5 h of remote tem-

3C

2C

IC

0

4 MARCH,1980

11 MARCH, 1980 . -

21 FEB., 1980

TEMPERATURE CALIBRATION

C =36.8%CD 33.9 C

I I I ~I I 0600 1000 1400 1800 2200

TIME (hours)

Fig. 6. Lidar temperature measurement calibration. A total of 8.5-hdata from different days are shown. Temperature data from a re-cording thermograph at the receiving telescope (solid line) were used

for the calibration of lidar temperature measurements.


10

DATA FROM THERMOGRAPH

- ABSOLUTE HUMIDITY

8 _

66- _, -.- :;:_

-24 _ _

0 __ __ I ___

0600 0700 0800 0900 1000TIME (hour)

Fig. 7. Lidar measurement of temperature and absolute humidityfrom the morning of 11 Mar. 1980. Error bars due to return signalfluctuations are shown for each data point. Temperature data are

compared with the thermograph recording.

perature data we collected during runs on three differ-ent days. The thermograph measurements are alsoshown. For all lidar temperature data the same cali-bration constant D was used. The best value for D was33.90C with a standard deviation of 2.30C. The majorfactor contributing to the error margin of the calibrationconstant is the inaccuracy of the etalon position. Im-proved calibration accuracy can be expected with betterfrequency selection accuracy.

Figure 7 shows the record of temperature and hu-midity from the morning of 11 Mar. 1980. Each pointrepresents an average of two stored data points, eachusing 150 shots/line. This results in a time resolutionof 90 sec/point. Error bars due to fluctuations of thereceived signal are indicated for each point. The out-side weather was cloudy at first, with rain starting at0740. The rain increased until it was so strong at 0810that the measurement had to be interrupted. At 0820the rain stopped, the cloud cover slowly dissipated, andthe wind increased. The remotely measured temper-ature is compared with a record of the thermograph atthe telescope. The general agreement is good. Somediscrepancies of the two measurements are expectedsince the remote measurement averages over a pathfrom 10 to 60 m above the ground and 775-m length,while the thermograph is located on the roof of abuilding. The discontinuities of temperature and hu-midity measurements at 0740, 0820, and 0905 h arecaused by realignment of the etalon in the OPO source.It is noticed that the drift of the transmitted frequenciescan cause considerable temperature errors.

The rms deviation for each temperature measure-ment is calculated from the return signal fluctuations.It varies with the meteorological conditions: in theearly morning, the SNR was 25. The uncertainty of thetemperature measurements from 0700 to 0730 h rangesbetween 1.10 C and 1.60 C. This is only a factor of 3larger than the value predicted by Eq. (6). Smallchanges of target reflectivity from pulse to pulse explainthe increase in noise.23 The SNR decreases later dueto the rain, and accordingly the rms deviations increaseto values between 1.60C and 2.00 C. After the rainstopped the SNR improved again to 25, and thetemperature uncertainty stayed between 1.50C and2.10 C. However, the measured values fluctuate morethan indicated by the error bars because of strongerhumidity fluctuations due to the wind. Any variationin absolute humidity during the 10-sec measurementcycle for each temperature point increases the uncer-tainty of this measurement. Shorter measurementcycles will improve the accuracy of the measurement.

Figure 8 shows a measurement taken on 21 Feb. 1980.This measurement run is interesting because a warmfront moved over Stanford at 2240 h. An increase intemperature can be observed in the lidar data as well asin the thermograph data. However, the remotelymeasured temperature jump is larger than that mea-sured at the telescope, as can be expected for measure-ments over an elevated path. Humidity data also show

C-3

I-

O L - 114 - ABSOLUTE HUMIDITY

12 .. .* ..-!..

4r-

2

0_______2100 2200 2300

TIME (hour)2400

Fig. 8. Lidar measurement of temperature and absolute humidityfrom the night of 21 Feb. 1980. Error bars due to return signal fluc-tuations are shown for each data point. Temperature data arecompared with the thermograph recording. During the measurement

a warm front passed over the measurement site.


an increase in water vapor concentration after passingof the warm front. It should be noted that the tem-perature data between 2210 and 2250 h are not reliable.A large error bar in both temperature and humidity dataindicates a disturbance at 2210 h. Afterward the errorbars of the temperature data are twice as large as before.Realignment of the OPO decreased the size of thetemperature uncertainties to the previous levels. Thecontinuation of the automatic measurement with themisaligned transmitter source is explained by the ab-sence of an operator.

IV. Discussion

We have demonstrated the capability of a contin-uously tunable infrared lidar to measure remotely pathaveraged temperature and humidity. Measurementswere done automatically under computer control.Selection of lines of the 1.9-Aum H20 band allows goodsensitivities. We achieved relative temperatureuncertainties of 1.50C for averages over 150 pulses. Therelative accuracy can be improved by more rapid tuningof the transmitter to eliminate errors due to changes ofthe target reflectivity and water vapor fluctuations.The absolute accuracy of the measurements of 2.30Cwas limited by the step size of the etalon tuning ele-ment.

The relative error of the humidity data is 1.5%. Theabsolute error, however, is estimated to 20% due touncertainties in the available cross-section data of theselected lines. More spectroscopic work is necessaryto improve these data.

Another noise source in the reported measurementsarises from variations of the target reflectivity betweenlaser pulses. Recent measurements show that topo-graphic targets such as atmospheric backscatteringsignals have temporal correlation times of the order ofa few milliseconds.2 3 The considerable improvementof relative accuracy should occur when all three probepulses are transmitted within this time. Ways toacomplish this with a single transmitter source areunder investigation.

The largest errors in the reported measurements aredue to inaccuracies in the frequency selection of theOPO transmitter. Improvement of the tuning mech-anism is possible and should lead to improved mea-surement accuracy.

Narrow linewidth operation of the OPO with an in-tracavity tilted etalon is possible but difficult due to therequirement for critical alignment and angle control.However, the OPO can be redesigned to avoid the needfor a tilted etalon. Initial studies show that OPO op-eration with a 75-grooves/mm echelle grating will allowoperation with a 0.15-cm-' linewidth without need fora tilted etalon.

In conclusion, we have carried out the first automatictemperature and humidity measurements using a con-tinuously tunable lidar system. The measurementsdemonstrated the potential for future depth resolvedtemperature and humidity measurements using a higher

energy tunable transmitter source. For DIAL mea-surements with a 150-m depth resolution at 1 km '500mJ must be transmitted using our present receiver/detector system. Improvements in detector sensitivity,telescope aperture, and source energy should allow re-mote temperature to 10C accuracy. Work is inprogress on a significantly higher energy tunable lasertransmitter that should meet the DIAL temperatureand humidity transmitter requirements.

We want to acknowledge the support of the U.S.Army through contract DAAG-77-G-0221 and the loanof one receiving telescope by the Electric Power Re-search Institute.

References1. L. Elterman, J. Geophys. Res. 58, 519 (1953).2. L. Elterman, J. Geophys. Res. 59, 351 (1954).3. R. G. Strauch, V. E. Derr, and R. E. Cupp, Appl. Opt. 10, 2665

(1971).4. J. J. DeLuisi, B. G. Schuster, and R. K. Sato, Appl. Opt. 14,1917

(1975).5. J. Cooney, J. Appl. Meterol. 11, 108 (1972).6. A. Cohen, J. A. Cooney, and K. N. Geller, Appl. Opt. 15, 2896

(1976).7. J. A. Cooney and M. Pina, Appl. Opt. 15, 602 (1976).8. R. Gill, K. Geller, J. Farina, J. Cooney, and A. Cohen, J. Appl.

Meterol. 18, 225 (1979).9. J. B. Mason, Appl. Opt. 14, 76 (1975).

10. H. I. Heaton, Astrophys. J. 212, 936 (1977).11. G. K. Schwemmer and T. D. Wilkerson, Appl. Opt. 18, 3539

(1979).12. E. R. Murray, D. D. Powell, and J. E. van der Laan, Appl. Opt.

19, 1794 (1980).13. J. E. Kalshoven, Jr., C. L. Korb, M. Dombrowski, and G.

Schwemmer, Abstract, Tenth International Laser Radar Con-ference, Silver Spring, Md. (1980), p. 68.

14. C. L. Korb and C. Y. Weng, Abstract, Ninth International LaserRadar Conference, Munich, Germany (1979), p. 185.

15. E. V. Browell, T. D. Wilkerson, and T. J. McIlrath, Appl. Opt. 18,3474 (1979).

16. R. A. McClatchey, W. S. Benedict, S. A. Clough, D. E. Burch, R.F. Calfee, K. Fox, L. S. Rothman, and J. S. Garing, "AFCRL At-mospheric Absorption Line Parameters Compilation," Envi-ronmental Research Paper 434, AFCRL-TR-73-0096 (Air ForceCambridge Research Laboratories, Bedford, Mass., Jan. 1973).

17. W. S. Benedict and L. D. Kaplan, J. Chem. Phys. 30, 388(1959).

18. R. L. Byer, Opt. Quantum Electron. 7, 147 (1975).19. R. M. Schotland, J. Appl. Meterol. 13, 71 (1974).20. M. Endemann and R. L. Byer, AIAA-81-0377, AIAA Nineteenth

Aerospace Sciences Meeting, St. Louis, Mo., 12 Jan. 1981.21. W. S. Benedict and R. F. Calfee, ESSA Professional Paper 2, U.S.

Department of Commerce, Washington, D.C., 1967.22. D. M. Gates, R. F. Calfee, D. W. Hansen, and W. S. Benedict, Nat.

Bur. Stand. U.S. Monogr. (1964), p. 71.23. N. Menyuk and D. K. Killinger, Postdeadline paper, Tenth In-

ternational Laser Radar Conference, Silver Spring, Md.(1980).


simultaneous remote measurements of atmospheric temperature and humidity using a continuously...

Documents