side-line tunable laser transmitter for differential absorption lidar measurements of co_2: design...

Side-line tunable laser transmitter for differentialabsorption lidar measurements of CO2: designand application to atmospheric measurements

Grady J. Koch,1,* Jeffrey Y. Beyon,2 Fabien Gibert,3 Bruce W. Barnes,1 Syed Ismail,1

Mulugeta Petros,4 Paul J. Petzar,5 Jirong Yu,1 Edward A. Modlin,1 Kenneth J. Davis,3

and Upendra N. Singh1

1NASA Langley Research Center, Hampton, Virginia, 23692, USA2Department of Electrical and Computer Engineering, California State University, Los Angeles, California 90032, USA

3Department of Meteorology, Pennsylvania State University, 415 Walker Building, University Park, Pennsylvania 16802, USA4Science and Technology Company, Hampton, Virginia 23666, USA

5National Institute of Aerospace, 100 Exploration Way, Hampton, Virginia 23666, USA

*Corresponding author: [email protected]

Received 14 November 2007; accepted 14 December 2007;posted 8 January 2008 (Doc. ID 89790); published 29 February 2008

A 2 μm wavelength, 90mJ, 5Hz pulsed Ho laser is described with wavelength control to precisely tuneand lock the wavelength at a desired offset up to 2:9GHz from the center of a CO2 absorption line. Oncedetuned from the line center the laser wavelength is actively locked to keep the wavelength within1:9MHz standard deviation about the setpoint. This wavelength control allows optimization of the op-tical depth for a differential absorption lidar (DIAL) measuring atmospheric CO2 concentrations. Thelaser transmitter has been coupled with a coherent heterodyne receiver for measurements of CO2 con-centration using aerosol backscatter; wind and aerosols are also measured with the same lidar and pro-vide useful additional information on atmospheric structure. Range-resolved CO2 measurements weremade with < 2:4% standard deviation using 500m range bins and 6:7 min (1000pulse pairs) integrationtime. Measurement of a horizontal column showed a precision of the CO2 concentration to < 0:7% stan-dard deviation using a 30 min (4500pulse pairs) integration time, and comparison with a collocated insitu sensor showed the DIAL to measure the same trend of a diurnal variation and to detect shorter timescale CO2 perturbations. For vertical column measurements the lidar was setup at the WLEF tall towersite in Wisconsin to provide meteorological profiles and to compare the DIAL measurements with the insitu sensors distributed on the tower up to 396m height. Assuming the DIAL column measurement ex-tending from 153m altitude to 1353m altitude should agree with the tower in situ sensor at 396m alti-tude, there was a 7:9ppm rms difference between the DIAL and the in situ sensor using a 30 min rollingaverage on the DIAL measurement. © 2008 Optical Society of America

OCIS codes: 280.1910, 140.0140, 280.0280, 010.3640, 010.3920, 010.7030.

1. Introduction

Increasing CO2 concentration in the atmosphereover the last century has been linked to climatechange, and future trends of CO2 will likely be of cri-

tical importance to management of carbon emissionsand sequestration. Improved capability for measur-ing CO2 concentration has been called for, especiallyactive remote sensing [1]. Studies have indicatedthat a precision of CO2 concentration measurementof 0.5% or smaller from a remote sensor is a goal forstudies of the global carbon cycle [2].0003-6935/08/070944-13$15.00/0

© 2008 Optical Society of America

944 APPLIED OPTICS / Vol. 47, No. 7 / 1 March 2008

Differential absorption lidar (DIAL) offers manyfeatures desirable for CO2 sensing in the technique’shigh spatial resolution, lack of dependence on ex-ternal light sources, relatively simple inversionmethods, and easy application to a variety of mea-surement geometries. Choice of the 2 μm wavelengthis based on absorption line parameters for absorptionstrength, insensitivity to temperature, freedom frominterference from other gases, and availability oflaser and optical technology [3]. Toward realizingpractical lidar systems, DIAL based on 2 μm wave-length lasers has shown progress toward high preci-sion measurements of CO2. Previous work at NASALangley Research Center demonstrated a measure-ment precision as low as 1% standard deviation forcolumn content and 2.5% for range-resolved mea-surements on 1000m long bins [4]. Independentwork at the Institut Pierre Simon Laplace, Labora-toire de Météorologie Dynamique, built another2 μm DIAL and calculated a precision as low as 1%for column content and established a high degreeof absolute accuracy [5]. We explore in this paperlaser technology advances to improve precisionfurther.An impediment to previous implementations of the

2 μm DIAL approach has been the lack of laserwavelength tunability and wavelength stability inrelation to the absorption line. Another DIAL instru-ment for water vapor profiling has shown the benefitof wavelength tuning for optimizing optical depth [6].In addition, analytical studies have confirmed theoptimization of DIAL by tailoring optical depth [7].

Beyond tunability, though, the wavelength must alsobe stabilized and held at its optimized value.

In previous work at NASA Langley Research Cen-ter, operation of the transmitter on-line wavelengthwas limited to the center of the absorption line. Theoff-line wavelength used was on the side of the linewith a small degree of tunability for selecting opticaldepth. This is, however, a disadvantageous approachfor two reasons. First, selection of both the on-lineand off-line wavelengths to be on the absorption fea-ture gives strong absorption that may limit rangecapability. Second, the choice of the off-line wave-length on the side of the line leads to complicationin knowledge of the differential absorption cross sec-tion as the absorption line changes shape with tem-perature and pressure. A better approach, describedin this paper, is to make the on-line wavelength ad-justable on the side of the absorption line (or center ifdesired) and to set the off-line wavelength well off ofthe absorption feature.

2. Lidar Design

The architecture of the transmitter, diagrammed inFig. 1, is a 90mJ, 140ns, 5Hz pulsed Ho:Tm:LuLiFoscillator injection seeded by continuous wave (CW)lasers to produce a single-frequency spectrum of3:4MHz linewidth. The pulsed laser has been de-scribed in detail in other publications and is onlybriefly discussed below in favor of concentrating onthe wavelength control implementation [8,9]. Useof the lutetium lithium fluoride (LuLiF) host materi-al is a recent development; our earlier work usedyttrium lithium fluoride (YLF). The use of LuLiF

Fig. 1. Layout of the lidar. Solid curves represent fiber optic paths. Dashed curves are electronic connections. Acronyms include AOM,acousto-optic modulator; EOM, electro-optic modulator; CW, continuous wave; and PZT, piezoelectric translator.

1 March 2008 / Vol. 47, No. 7 / APPLIED OPTICS 945

or YLF involves several trade-offs. The advantage ofLuLiF is that it offers 20% higher pulse energy thanYLF. The advantage of YLF is that the absorptionline it matches near 2051nm is somewhat less sen-sitive to temperature and pressure than the LuLiFline at 2053nm. There is also a difference (the LuLiFline is stronger) in absorption strength between theYLF and the LuLiF lines, which could affect DIALperformance if operation were limited to the absorp-tion line center, but with side-line tunability the ab-sorption line strength is not as critical a factor. Thecapability to double pulse the laser in which the laseris fired twice per pump cycle is different between thetwo lasers and is a subject under further study [9].The DIAL measurements described here were madewith a single-pulse format, alternating the wave-length among pulses at 5Hz pulse repetition fre-quency. LuLiF was used in this study, a choicedriven primarily by the alternate application ofthe laser transmitter for wind measurements inwhich high pulse energy is desired [10].The pulsed laser is injection seeded by CW lasers,

commercially available Ho:Tm:YLF designs produ-cing approximately 50mW in a single-frequencyspectrum. These lasers can be coarsely tuned over∼5nm by selection and tilting of an intracavityetalon. Once the etalon is adjusted to tune the outputwavelength near an absorption line of interest, finetuning is accomplished with piezoelectric translator(PZT) motion of the output coupler. Three CW lasersare in use and named for their wavelength setting: acenterline reference, tunable side line, and off line;Fig. 2 shows the wavelength settings of these lasersin relation to the CO2 absorption line. Phillips et al.developed a similar technique for control of a CWlaser absorption spectrometer but referenced to a dif-ferent absorption line [11]. In this paper we have theadded complexity of using the CW lasers to injectionseed a pulsed laser.The centerline reference serves as the master fre-

quency control referenced to a gas cell. It is locked

onto the line center by a frequency modulation (FM)spectroscopic technique, similar to a wavelengthmodulation (WM) technique described previouslyto stabilize an Ho:Tm:YLF laser [12]. The differencebetween the FM and theWM techniques is in the fre-quency of modulation of the laser; FM typically refersto modulation at radio frequencies and WM to audiofrequencies. In general, the FM technique can be ex-pected to offer a better level of stabilization, becausethe frequency used involves less 1=f noise. We founda stabilization level over long time scales (manyhours) to within a standard deviation of 370kHz ofthe line center, approximately a factor of 10 betterthan our experiments with WM. We chose an FM fre-quency of 175MHz with regard to matching the line-width of the CO2 line contained in a gas cell at∼5Torr pressure. This modulation is applied withan electro-optic phase modulator and offers anotheradvantage over WM in that, by picking off the laserbeam before it enters the phase modulator, accesscan be had to an unmodulated spectrum, WMis usually accomplished by modulating the outputcoupler PZT resulting in a modulated spectrumthroughout the beam’s path. Having access to the un-modulated beam is an important feature for the nextstep in establishing the tunable side line. A note forfuture research is that the centerline reference neednot be a solid-state laser and could be replaced with adiode laser that requires less power and less cost. Forexample, Mitsuhara et al. showed a distributed feed-back diode laser that could serve as a centerlinereference [13]. The disadvantage of the diode lasercompared with the solid-state Ho:Tm:YLF is thatthe diode laser is limited in output power to less than10mW. The centerline reference does not requiremuch power (5mW), and a diode laser would besuitable. The other two lasers, however, requirehigher power of a solid-state laser for injection seed-ing the pulsed laser.

The tunable side-line wavelength is generated byoffsetting an Ho:Tm:YLF laser from the centerlinereference. This offset is created by heterodyningthe two lasers together and set at a desired off-set by a proportional-integral-differential feedbackcontrol loop. A similar scheme, but not involvingreferencing to an absorption line, has been demon-strated by Hale et al. for Doppler wind lidar in whicha frequency agile local oscillator was required [14].The offset can be selected between 0.1 and 2:9GHz.Figure 3 shows the results of characterizing the per-formance of the offset lock set for 1:2GHz frequencydifference. With the lock applied an 870kHz stan-dard deviation in the frequency difference was foundover long (hours) time scales. Also shown for compar-ison is the offset frequency without the lock appliedafter setting the frequency difference at 1:2GHz,demonstrating the need to have active control. Itshould be noted that the control loop maintains onlythe frequency difference and has no way to determineto which side of the line center the frequency offset isbeing maintained; this determination was made in

Fig. 2. Wavelength settings of CW lasers superimposed on aHITRAN database simulation of transmission over a 1km horizon-tal path at standard temperature and pressure. The side-line laseris shown here at three wavelength settings used in creating thedata of Figs. 4 and 5.


our setup by verifying the side-line wavelength witha wavemeter. Only a small part (less than 1mW) ofthe side-line laser is used for wavelength control;most of its power is sent to the pulsed laser for injec-tion seeding.The third CW laser in use is tuned well off the ab-

sorption line and left without any active wavelengthcontrol. The measured long term (hours) wavelengthdrift was found to be 7pm, not enough to create sig-nificant variations in determining the differentialabsorption cross section for DIAL calculations. Theoff-line laser and side-line laser are both coupled intoan electro-optic switch for alternating which wave-length is passed to the pulsed laser. This switchcan make a transition in < 1 μs, making it suitablefor a format in which the pulsed laser is operatedto fire a doublet of pulses on each pumping cycle witheach pulse of the doublet seeded at a different wave-length (a laser design for future study). In the imple-mentation described here, the pulsed laser was in asingle-pulse format with on-line and off-line wave-lengths alternating at the 5Hz pulse repetition fre-quency of the laser.An important aspect for the CO2 measurements is

the knowledge of the side-line wavelength. It hasbeen shown that there are two wavelength controlloops in the centerline reference and the side-lineoffset. There is another source of wavelength uncer-tainty, however, in the control loop to injection seedthe pulsed laser. This control is established by aramp-and-fire technique, and it has been verifiedby using this technique that the pulsed laser willclosely follow the injection seeded wavelength [15].While the ramp-and-fire arrangement does allowthe pulsed laser to match the injection seed, there

can be some residual frequency jitter introduced.This jitter has been extensively characterized andis indeed a crucial consideration in the use of sucha laser transmitter design in coherent lidar. In a co-herent lidar, which is the lidar technique used in theatmospheric tests described later, the frequency jit-ter is measured on a pulse-to-pulse basis. A typicallevel of jitter was found to be 1:61MHz standard de-viation. To evaluate the overall frequency knowledgeand stability of the side-line wavelength, we mustconsider that three locking loops are involved. Thesquare root of the sum of the variances (370kHzfor the centerline reference, 870kHz for the side-lineoffset locking, and 1:61MHz for the ramp-and-fire in-jection seeding) of each locking loop gives an overallwavelength stability of 1:9MHz standard deviation.This residual frequency uncertainty was calculatedto present an error of less than 0.1% of the DIALmeasurement, much less than other potential errorsources such as uncertainty in temperature, pres-sure, and humidity [16].

Before reaching the pulsed laser, the injectionseeded beam path is frequency shifted by 105MHzwith an acousto-optic modulator (AOM) to createan intermediate frequency for coherent detection.No measurable jitter was found from the AOM,but it does create a frequency offset that must be ac-counted for in figuring the differential absorptioncross section. Use of the AOM also creates a pointat which to split off a portion of the CW beam touse as a local oscillator. The zeroth order beamthrough the AOM is taken for this purpose.

The output of the pulsed laser is directed toward a4 in: (10 cm) diameter telescope and associatedheterodyne photoreceiver, identical to the design re-ported in a previous paper [10]. The data acquisitionand signal processing system is likewise similar, withthe addition of refined signal processing techniquesdescribed in detail in another paper [17]. To summar-ize the data acquisition and signal processing, it be-gins with digitization of the heterodyne signal at500Ms=s with 8 bits. Consecutive pulses of a numberdepending on measurement application are acquiredwith the wavelength alternating among pulses.

Later processing of the acquired data is carried outby LabVIEWalgorithms on a personal computer. Thesequence of pulses is first separated into on-line andoff-line pulses. Each pulse is divided into range binsand Fourier transformed before averaging in the fre-quency domain. The energy of the average spectra ofeach range bin is calculated by first estimating thenoise floor by a technique of minimizing mean-square error; a noise model is created with a firstorder polynomial as a function of frequency. Afterdefining the noise floor, a threshold is set abovewhich spectral samples are integrated to determineenergy in a range bin.

3. Atmospheric Experiment 1: Varying Optical Depth

To explore application of the side-line tuning techni-que, atmospheric measurements were made with

Fig. 3. Characterization of offset locking between centerline andside-line lasers. For comparison the locking electronics were alsodisengaged and the side-line laser was left to drift; a drift of tens ofmegahertz is typical over a 1h long time span. The discontinuousjumps in the drifting laser are attributed to mechanical perturba-tions in the laboratory environment such as closing a door.


different side-line tuning frequencies. In this experi-ment the lidar beam was pointed vertically, andDIAL measurements were made at side-line offsetfrequencies of 1:7GHz (24pm), 2:2GHz (31pm),and 2:7GHz (38pm) from the absorption line center.For each frequency setting, 2000 pulses were ac-quired with the on-line wavelength alternating withthe off-line wavelength. To minimize atmosphericvariations, the time between DIAL measurementswas kept as short as possible with ∼2 min betweenmeasurements required to change the offset lockfrequency.Figure 4 shows the power of the atmospheric back-

scatter for the three on-line wavelengths and the off-line wavelength processed on 200m long range bins.As expected, the backscatter signal drops morerapidly for wavelength settings closer to the line cen-ter. Atmospheric aerosol structure can be seen withhigher backscatter in the atmospheric boundarylayer under 1400m altitude compared with the at-mosphere above 1400m. The data were recorded∼1 h after sunset on 1 August 2006, so the aerosollayer to 2800m above the boundary layer is likelya residual from the higher daytime boundary layer.Inspection of the off-line signal at 2800m altitudeand beyond shows that a small signal extends higherinto the free troposphere, but with the extra absorp-tion the on-line signals do not show this signal.Analysis of the DIAL measurements proceeds with

calculating the optical depth as a function of altitudefrom the data of Fig. 4. Optical depth is defined as

τ�R� � 12ln

Soff �R�Son�R�

; �1�

where Son and Soff are the energies of the on- and off-line signals and R is a range. Figure 5 shows the re-sults of this calculation for the three cases of wave-length detuning from the line center. In comparing

the three cases of the optical depth, the slope ofthe optical depth is seen to be higher the closerthe wavelength setting is to the line center. Thisdemonstrates the utility of the side-line tuning tech-nique to set a desirable optical depth to balancestrong enough differential absorption for sensitiveDIAL measurements but not too much absorptionto prevent measurements at distant ranges. The ef-fects of low signal-to-noise ratio (SNR) from absorp-tion are seen in Fig. 5 as the optical depth becomesbiased with low SNR, then becomes nonsensicalas the SNR becomes unity. Analysis of the biaswith SNR indicates in the case of Fig. 5 thatgood DIAL measurements could be made to2200m for 1:7GHz detuning, 2600m for 2:2GHzdetuning, and 3000m for 2:7GHz detuning.

Also evident in Fig. 5 is that there is a change inoptical depth between the boundary layer below1400m and above. This change could be related toa different CO2 concentration in the boundary layerand free troposphere or, more likely, related to a dif-ference in temperature, pressure, and humidity be-tween the two layers. To convert optical depth toCO2 concentration these meteorological parametersmust be known. No such ancillary measurementswere made in this test of changing optical depthbut are included in the atmospheric tests describedin Sections 4 and 5.

4. Atmospheric Experiment 2: Precisionand Sensitivity

Assessment was made of the precision of the DIALresults by repeatedly making CO2 measurementsin as short a time span as possible and analyz-ing the statistics of the resulting CO2 data pro-duct assuming the CO2 concentration does notchange over the measurement period. That is, a

Fig. 4. Backscatter signal power at four wavelengths: off-line andthree side-line wavelengths at different offsets. Fig. 5. Optical depth for three wavelength offsets.


DIAL calculation was made of CO2 number density,n, from

n � 12Δσ�R2 − R1�

ln�Son�R1�Soff �R2�Son�R2�Soff �R1�

�; �2�

where Rx is a range andΔσ is the differential absorp-tion cross section; Δσ is a function of temperatureand pressure, so the lidar beam was pointed horizon-tally in these tests to avoid variation of temperatureand pressure with altitude. Aweather station locatednear the lidar provided temperature and pressurethat were later input into analysis of the CO2 lineshape to calculate Δσ. Absorption line parametersto make this calculation were based on the HITRANdatabase, modified using refined measurementsof the absorption line reported by Regalia-Jarlotet al. [18]. In addition to calculating absorption lineparameters, temperature and pressure data are usedto normalize the number density of Eq. (2) to find adry mixing ratio. Six sets of 2000pulse pairs weretaken, requiring 6:7 min to acquire each set. In ad-dition to unaccounted for instrument parametersthat are desired to be characterized, several errorsources can influence this calculation that are diffi-cult to quantify including fluctuations of CO2 concen-tration over the measurement period, fluctuations intemperature and pressure not accurately repre-sented by the weather station located 1km away,and variations in aerosol backscatter over the 200msbetween on-line and off-line pulses. With the diffi-culty of characterizing these atmospheric variables,the measured precision represents an upper bound

on the combination of many error sources. The in-strumental error should be considered as less thanthe combined value. Experimental results showeda precision of < 2:4% standard deviation of the meanusing a 2:7GHz side-line offset, 500m long rangebins, and 6:7 min integration time.

This precision can be improved in several ways.First, the range bin size could be increased if lessrange resolution is tolerable. Second, the integrationtime could be increased. We chose a 6:7 min integra-tion time because this length is manageable forstorage by the data acquisition system. However,CO2 variations are expected to occur over longer timescales with a 30 min integration time indicated to beacceptable [2]. With integration of 30 min the preci-sion found for 6:7 min could be reduced by a factor of2 to a level of < 1:2%. Third, the pulse repetition rateof the laser could be increased to put more pulses intothe measurement period. While the current instru-ment operates at a 5Hz pulse repetition rate, designsare under way to increase to a 10Hz rate using a si-milar laser architecture. Recent advances in usingTm fiber lasers at 1:9 μm output to pump an Ho so-lid-state laser show promise in producing pulse repe-tition rates in excess of 100Hz [19].

A horizontal column measurement is convenientfor comparing with an in situ sensor, and such a sen-sor was placed next to the lidar for such a study. Thein situ sensor was a modified Li-Cor nondispersiveinfrared gas analyzer and has been extensively usedand calibrated in many field and airborne studies[20,21]. The in situ sensor inlet was placed ∼5maway from, and at the same 4m height of, the lidarbeam exit point. Ideally, the in situ sensor would be

Fig. 6. Line-of-sight wind speed (top) and backscatter signal power (bottom) on the morning of 23 March 2007. Off-line pulses are takenfrom the DIAL data record and grouped into 20pulse sets for processing here. One vertical stripe here thus represents an integration timeof 8 s.


placed in the middle of the column that the lidar isprobing, but the complex topography of the measure-ment site made this impossible. Topography alsolimited how close to a horizontal angle the lidar couldview; trees at a 2km range required a 0:5° eleva-tion above true horizontal. The terrain over whichthis column was taken was complex, passingover a grassy field, a marsh, a wide creek, and awooded area.As shown in Figs. 6 and 7 lidar measurements

were made continuously from 5:58 a.m. to 12:02 p.m.on 23March 2007 with the intent of probing CO2 var-iations that can occur in night-to-day transitions. InFig. 6 the lidar pulses have been grouped in 20pulsesets for averaging to show backscatter signal powerand line-of-sight wind speed. Wind measurement,made by analyzing the frequency shift of the aerosolbackscatter, is a benefit of using a heterodyne detec-tion receiver and gives a dual use for the lidar. Theaerosol backscatter shows a typical inverse range-squared decay out to a 4:2km range, where the acqui-sition record length ended. Looking at the trend ofthe backscatter it is seen to be decreasing in the latemorning hours, likely a consequence of convectivelifting spreading aerosols through a higher altitudeas the daytime atmospheric boundary layer forms.

The wind signal shows movement ranging from 3to 8m=s away from the lidar. The overall increasein speed with range is likely associated with theslight elevation angle of the lidar beam, such thatthe beam at further ranges is at a higher altitudeand encountering a higher wind speed. There is alsoa change in the wind at the 1200m range, where thespeed drops; this is probably an effect of topography,and this range corresponds with the distance towhere a wide creek borders a wooded area.

For DIAL measurements 2000 pulses weregrouped for processing on 500m range bins. Analysisof the on-line SNR shows that unbiased DIALmeasurements could be made to a range of 1903m,giving three 500m range bins to work with. Sincethe lidar is oriented to probe horizontally, and adopt-ing the assumption that the CO2 concentration isuniform in the horizontal direction, we averagedthe measurements over these three range bins. A2000 pulse set represents an integration time of6:7 min, and four of these sets will be applied toform a moving average. With four 2000pulse setsaveraged and three range bins averaged, the preci-sion of the DIAL measurements is estimated tobe < 2:4%=sqrt�12� � 0:7%.

Fig. 7. DIAL results (blue) compared to in situ sensor (green) of 23 March 2007 from Hampton, Virginia. The DIAL beam is pointinghorizontally for a columnmeasurement at the same altitude as the in situ sensor. The in situ sensor is an infrared gas analyzer that drawsin a sample of air to test; it provides a measurement of CO2 concentration at 1 s intervals. The sharp peaks in the in situmeasurement arecaused by CO2-rich pockets of air traveling through the test site. The DIAL results in blue are shown using a 30 min rolling average; thosein red are a 6:7 min integration. The DIAL shows both excellent performance in accuracy and precision. Both sensors show an approximaterise of 8ppm before sunrise, followed by an ∼10ppm decline. The short term peaks between the two sensors cannot be strictly compared,because the two sensors are probing different volumes of air. However, the DIAL is picking out short term variations. The DIAL’s precisionin these measurements is calculated at < 0:7%.


The DIAL and in situ results are shown in Fig. 7.The in situ sensor making measurements on a 1 s in-terval shows short time scale (minutes) variations ofCO2 of as much as 15 ppm as pockets of CO2-rich airtravel through the sampling volume. Even in night-time conditions, before convection initiates, there areshort time scale variations of 5ppm; these fastchanges in CO2 cast doubt on the assumption of uni-form CO2 concentration over a 40 min measurementtime used in calculating DIAL precision and suggestthat the instrument precision is indeed better thanthe calculated upper bound of 0.7%. The in situ sen-sor also reveals the trend over several hours of CO2increasing in the nocturnal boundary layer beforesunrise, then a decrease in CO2 concentration aftersunrise when photosynthetic CO2 uptake initiatesand sunlit convection begins.The DIAL results show good agreement with the

in situ sensor, revealing the same trend beforesunrise of increasing CO2, then decreasing CO2.The DIAL also measures perturbations on timescales of < 1h of the same amplitude as, but dis-placed in time and not of as high a time resolu-tion, of the in situ sensor. Short time scale variationsare not expected to match in the two sensors, becausethey are probing different volumes of air: a singlepoint for the in situ sensor and a 1903m long col-umn for the DIAL. The agreement between the twosensors can be quantified if an assumption is madethat over the 7h course of the measurement that themean value should be the same. The in situmeasure-ment was rolling averaged over a 30 min window forthis assessment to match the lidar. The in situ sensorwas found to give a mean value of 401:53ppm(σ � 2:46ppm), and the lidar was found to give amean of 402:44ppm (σ � 4:69ppm), a difference inthe mean of 0:91ppm and indicating a bias of thelidar measurement of 0.2% assuming the in situsensor as reference truth.The DIAL demonstrates here a high precision and

sensitivity, enough to well represent a small diurnalvariation. Diurnal variation can easily be larger thanthe 10ppm revealed here on an early spring morn-ing; summertime variations through sunrise canexceed 50ppm when convection is strong and photo-synthetic activity is high.

5. Atmospheric Experiment 3: Vertical Viewing

Vertically viewing measurements are more difficultthan horizontally viewing measurements to compareagainst an in situ sensor, because the in situ sensormust be positioned at a high altitude, at least higherthan the 150m minimum range of the lidar. In addi-tion, meteorological profiles of temperature, pres-sure, and humidity must be known to calculate thedifferential absorption cross section and to reducethe DIAL measurements to a CO2 dry mixing ratio.To provide for these needs the lidar was transportedto the site of the WLEF tower in the Chequamegon-Nicolet National Forest of northernWisconsin. Whilethe WLEF tower is primarily a television and radio

transmitter operated by the Wisconsin EducationalCommunications Board, it has been outfitted within situ CO2 sensors and meteorological sensors sup-porting carbon cycle studies dating back to the mid1990s [22,23]. During this study, in situ CO2 sensorswere located at heights of 11, 30, 76, 122, 244, and396m. The lidar was positioned underneath thetower, ∼40m away from the tower’s center line.

The tower in situ CO2 measurements for a 24hperiod starting on the evening of 13 June 2007 areshown in Fig. 8. Weather conditions over this timewere of clear skies with occasional cumulus cloudsduring daytime hours. A CO2 diurnal cycle at differ-ent heights is illustrated by these in situ measure-ments. At lower altitudes, CO2 builds up duringnighttime hours as respiration is trapped withinthe nocturnal boundary layer. The buildup of CO2is less as altitude increases, with measurements at396m altitude showing no nocturnal CO2 buildupas this altitude is above the nocturnal boundarylayer height. After sunrise, convection begins tospread the CO2 trapped at low levels to higher alti-tudes. Throughout this 24h time the CO2 concentra-tion at 396m shows only a few parts per millionvariation around a value of 373ppm. These in situmeasurements suggest that the lidar should mea-sure little variation as well, since the lidar startsmeasuring a column starting at a 150m altitude thatextends to beyond 1km. While the tower in situmea-surements at 244m (within the column measured bythe lidar) show an increase in CO2 around 9 a.m.peaking at 8ppm, this perturbation will likely notbe seen by the lidar as the more constant CO2 con-centration of higher altitudes of the column will aver-age out the small change in a small section at thebeginning of the column.

Fig. 8. In situ CO2 measurements of 13–14 June 2007 at theWLEF tower. During nighttime hours soil respiration releasesCO2 that becomes trapped in the nocturnal boundary layer. Asthe Sun rises the trapped CO2 is released to higher altitudes byconvective mixing, and photosynthetic activity reduces the CO2

concentration. The sensor at 396m altitude shows little variationthroughout this time period, suggesting that the DIAL should alsomeasure little variation.


To support the DIAL calculations meteorologicalprofiles were made based on sensors of temperature,pressure, and humidity located on the tower. Sincelidar measurements extend to higher altitudes thanthe tower represents, the tower measurements wereextrapolated to several kilometer altitude. DIAL cal-culations were made on 200m range bins and thenweighted with the meteorological calculations atthe corresponding altitude. With the assumption es-tablished by the data of Fig. 8 that CO2 concentrationis constant with altitude, the measured CO2 concen-tration in multiple range bins, to an altitude to whichusable SNR persisted, were averaged together torepresent a column from 153 to 1353m altitude.Longer range bins, such as the 500m long bins usedfor horizontal column measurements, are desirablefor higher precision but can create more bias as alti-tude resolution is lost in compensating for meteoro-logical variables; 200m bins were thus selected as acompromise between precision and meteorologicalrepresentation.The DIAL column measurements are shown in

Fig. 9, taken over the same time period of the towerin situ data of Fig. 8. An integration time of 2 min(600pulses) was used and then rolling averaged over30 min. Comparison of the 30 min DIAL results withthe in situ measurements at 396m show that theDIAL is measuring around the same value of373ppm, but the DIAL results are noisier than thecase of horizontal columnmeasurements. This addednoise is attributed to the more rapid loss of signalwith range in the vertical viewing case due to the de-creasing concentration of aerosols encountered invertical viewing. Vertical viewing measurementscan also be noisier than horizontal measurementdue to the complexity of compensating for meteoro-logical variables’ variation with altitude. If the as-sumption is made that the column measurement

results should be the same as the tower in situ mea-surement at 396m altitude, then there is a 7:9ppmrms difference between the in situ sensor and theDIAL measurement rolling averaged over 30 min.If the DIAL is averaged over 1h then the rms differ-ence is 6:3ppm. These DIAL calculations were madebased on a traditional approach of binning the atmo-spheric backscatter into fixed widths. An alternateapproach of fitting the slope of the measured opticaldepth to find CO2 concentration as described byGibert et al. is under evaluation [5].

The DIAL measurements also contain informationon aerosol content and wind. Since the lidar beamwas pointing vertically during the DIAL measure-ments, the wind measurements are of the verticalcomponent of wind. Vertical wind features can occuron time scales shorter than the 2 min integrationperiod of the DIAL measurements, so the DIALmeasurements were broken up into smaller setsfor analyzing wind. Furthermore, the off-line tunedpulses are of more interest since they are not attenu-ated by absorption and reach to higher altitudes. Forthese considerations the off-line pulses were groupedinto sets of 20pulses (representing an integrationtime of 8 s) for processing. The results of this proces-sing are shown in Fig. 10 in three time segments eachspanning 1h 20 min representing evening, night,and morning. Vertical wind wave and turbulentstructure can be seen to be tapering off in the eveninghours, very weak in the nighttime section, and verystrong in the late morning. The backscatter signalpower shows variation throughout the day as theatmospheric boundary layer changes; the nighttimesignal power near 1km altitude is about 5dB lessthan the morning hours.

Aside from vertical wind measurements coincidentwith zenith-viewing DIAL, the horizontal wind vec-tor was measured every half hour during the DIALset. The wind profiles, shown in Fig. 11, were madewith the lidar beam pointed in three different direc-tions to take measurements of the horizontal andvertical wind. The first two directions were madeat a 30° elevation angle and orthogonal azimuthsto find the two components of the horizontal windvector. The vector sum of these two components isshown in the first two panels of Fig. 10 as the hori-zontal wind speed and horizontal wind direction. Thethird look direction was taken with the beam atzenith to measure vertical wind. The vertical windspeed and backscatter power form the lower twopanels of Fig. 11. At each look angle 20pulses wereaccumulated for averaging; with time required tomove the beam scatter to each orientation the totalamount of time for the wind profile was ∼30 s. Mea-surement of the horizontal wind profile reveals addi-tional structure of the atmosphere including anocturnal jet, wind shear, and variation of the windvector with altitude. Wind data combined with theCO2 concentration can be used to determine thetransport of CO2 in horizontal and vertical direc-tions, a subject of ongoing study.

Fig. 9. DIAL and in situmeasurements of 13–14 June 2007 at theWLEF tower site. The in situ sensor is at 396m altitude, the high-est sensor from Fig. 8. The DIALmeasurements are over a verticalcolumn extending from 153 to 1353m altitude.


6. Conclusion

A 2 μm wavelength single-frequency tunable lasertransmitter has been built suitable for DIAL mea-surements. The tunability of the laser, along withthe ability to lock the wavelength, allows selectionof a position on the CO2 absorption line to optimizeoptical depth and thereby optimize the sensitivityand precision of the DIALmeasurements. Character-ization of the wavelength lock showed stabilizationto within 1:9MHz standard deviation of the pro-grammed wavelength setting, including jitter effectsfrom three control loops: lock to centerline, lock of theside-line offset, and lock of the pulsed laser to the in-jection seed. The wavelength offset from theabsorption line center can be as much as 2:9GHz.The newly developed transmitter was built into a

complete lidar system using a coherent heterodynereceiver. Variation of the laser wavelength setting

from the line center showed that the optical depthcan be changed. Precision of DIAL measurementswas assessed by repeating measurements as quicklyas possible and analyzing the statistics of the concen-tration measurement, assuming that the CO2 con-centration does not change over the measurementtime. A precision of less than 2.4% standard devia-tion was found for 500m long range bins and anintegration time of 6:7 min (1000 pulse pairs). Preci-sion can be improved by a longer integration time orby averaging multiple bins together with the as-sumption that CO2 concentration does not changeamong bins. For example, a horizontal column wasmeasured using 30 min integration and three500m range bins averaged together, with a resultingprecision of less than 0.7%. Such horizontal columnmeasurements were made over many hours to cap-ture part of a diurnal cycle and compared with an

Fig. 10. Continues on facing page.


Fig. 10. Vertical backscatter and wind measurements derived from off-line pulses during DIAL measurements. Three times of day arerepresented: (a) 6:53 to 8:11 p.m., (b) 2:00 to 3:18 a.m., and (c) 10:36 to 11:54 a.m. While the integration time for a DIALmeasurement was2 min, the vertical wind measurements have been separated into 8 s long integration periods to better resolve vertical wind features. Thetop panel of each set is the vertical wind; the lower plot is backscatter signal energy. Vertical motion can be seen to be moderate (a) in theevening hours, (b) dissipating in the early morning, and (c) very strong in late morning. The trend in aerosol density can also be seen with aminimum during the early morning hours.

Fig. 11. Wind profile sampled every half hour on 13–14 June 2007. The panels of data show, from top to bottom: (1) horizontal wind speed,(2) horizontal wind direction, (3) vertical wind speed (red upward, purple downward), and (4) backscatter signal power. The horizontalwind speed shows a nocturnal jet peaking at 10m=s. A half hour sampling of vertical wind is too coarse to reveal thermals and wave action,hence the processing shown in Fig. 10. Variations of backscatter signal power show a decreasing trend in nighttime hours followed by anincrease as daytime convection lifts aerosols.


in situ sensor placed at the same height as the DIALbeam. Close agreement was found between the insitu sensor and the DIAL column.For characterization of the lidar performance in a

zenith-viewing operation, the lidar was transportedto the site of the WLEF tower in northern Wisconsin,on which CO2 and meteorological sensors are placedup to a height of 396m. The lidar was set with thebeam viewing toward zenith for DIAL and scannedevery half hour for wind profiling. With the in situsensor at 396m altitude showing little variationaround 373ppm, the DIALmeasurements over a ver-tical column extending from 153 to 1353m were alsoexpected to measure little variation. The DIAL mea-surements indeed showed the same average concen-tration of 373ppm. The difference between the in situsensor and the DIALwas 7:9ppm rms using a 30 minintegration time on the DIAL. Horizontal and verti-cal wind profiles made by the lidar over the sametime as the DIAL were demonstrated and give addi-tional meteorological information.Future work includes evaluating improvements in

laser and optical components. This includes testingan alternate receiver design using direct detectionand a new development in photodetector technology,an InGaSbAs/AlGaAsSb heterojunction phototran-sistor [24]. Direct detection, if a photodetector can of-fer the same SNR as heterodyne detection, canprovide higher precision DIAL concentration mea-surements, since direct detection avoids speckle byaperture averaging. The reason for not previouslypursuing direct detection has been the lack of photo-detectors of sufficient detectivity. Another approachto improving precision is to average more pulses witha higher pulse repetition rate transmitter. Designsare under way to extend the pulse repetition rateof the transmitter from 5 to 10Hz, and possibly touse a double pulse format.

The authors thank Roger Strand and Jeffrey Ayersof the Wisconsin Educational CommunicationsBoard for hosting the lidar at the WLEF tower.Yonghoon Choi of the National Institute of Aerospaceand Stephanie Vay of NASA Langley Research Cen-ter provided the in situ measurements for the hori-zontal column comparisons. Arlyn Andrews of theNational Oceanic and Atmospheric Administration(NOAA) Earth System Research Laboratory pro-vided the in situ CO2 data from the WLEF tower.This research was funded by the NASA InstrumentIncubator Program and the NASA Laser Risk Reduc-tion Program.

References1. Space Studies Board, National Research Council, Earth

Science and Applications from Space: National Imperativesfor the Next Decade and Beyond (National AcademiesPress, 2007).

2. S. Ismail, G. Koch, N. Abedin, T. Refaat, M. Rubio, K. Davis, C.Miller, S. Vay, and U. Singh, “Development of a ground-based2-micron differential absorption lidar system to profile tropo-

spheric CO2,” in NASA Earth Science Technology Conference2006, paper B9P3.

3. R. T. Menzies and D. M. Tratt, “Differential laser absorptionspectrometry for global profiling of tropospheric carbon diox-ide,” Appl. Opt. 42, 6569–6577 (2003).

4. G. J. Koch, B. W. Barnes, M. Petros, J. Y. Beyon, F. Amzajer-dian, J. Yu, R. E. Davis, S. Ismail, S. Vay, M. J. Kavaya, andU. N. Singh, “Coherent differential absorption lidar measure-ments of CO2,” Appl. Opt. 43, 5092–5099 (2004).

5. F. Gibert, P. H. Flamant, D. Bruneau, and C. Loth, “Two-micrometer heterodyne differential absorption lidar measure-ments of the atmospheric CO2 mixing ratio in the boundarylayer,” Appl. Opt. 45, 4448–4458 (2006).

6. A. S. Moore, K. E. Brown, W. M. Hall, J. C. Barnes, W. C.Edwards, L. B. Petway, A. D. Little, W. S. Luck, I. W. Jones,C. W. Antill, E. V. Browell, and S. Ismail, “Development ofthe lidar atmospheric sensing experiment (LASE). An ad-vanced airborne DIAL instrument,” in Advances in Atmo-spheric Remote Sensing with Lidar, A. Ansmann, R. Neuber,P. Rairoux, and U. Wandinger, eds. (Springer-Verlag, 1996),pp. 281–288.

7. D. Bruneau, F. Gibert, P. H. Flamant, and J. Pelon, “Comple-mentary study of differential absorption lidar optimization indirect and heterodyne detection,” Appl. Opt. 45, 4898–4908(2006).

8. J. Yu, B. C. Trieu, E. A. Modlin, U. N. Singh, M. J. Kavaya,S. Chen, Y. Bai, P. Petzar, and M. Petros, “1J=pulse Q-switched 2 μm solid-state laser,”Opt. Lett. 31, 462–464 (2006).

9. J. Yu, A. Braud, and M. Petros, “600mJ double-pulsed 2-micron laser,” Opt. Lett. 28, 540–542 (2003).

10. G. J. Koch, J. Y. Beyon, B. W. Barnes, M. Petros, J. Yu, F. Am-zajerdian, M. J. Kavaya, and U. N. Singh, “High-energy 2 μmdoppler lidar for wind measurements,” Opt. Eng. 46, 116201(2007).

11. M. W. Phillips, J. Ranson, G. D. Spiers, and R. T. Menzies, “De-velopment of a coherent laser transceiver for the NASA CO2

laser absorption spectrometer,” in Proceedings of the 2004Conference on Lasers and Electro-Optics, San Francisco,Calif.

12. G. J. Koch, A. N. Dharamsi, C. M. Fitzgerald, and J. C.McCarthy, “Frequency stabilization of an Ho:Tm:YLF laserto absorption lines of carbon dioxide,” Appl. Opt. 39, 3664–3669 (2000).

13. M. Mitsuhara, M. Ogasawara, M. Oishi, H. Sugiura, andK. Kasaya, “2:05-μm wavelength InGaAs-InGaAs distribu-ted-feedback multiquantum-well lasers with 10mW outputpower,” IEEE Photon. Technol. Lett. 11, 33–35 (1999).

14. C. P. Hale, S. W. Henderson, and D. M. D’Epagnier, “Tunablehighly-stable master/local oscillator laser for coherent lidarapplications,” in Proceedings of Tenth Biennial Coherent LaserRadar Technologies and Applications Conference, Mt. Hood,Oregon, 28 June 1999 (Universities Space Research Associa-tions, Huntsville, Ala., 1999), pp. 115–118.

15. G. J. Koch, M. Petros, J. Yu, and U. N. Singh, “Precise wave-length control of a single-frequency Ho:Tm:YLF laser,” Appl.Opt. 41, 1718–1721 (2002).

16. F. Gibert, P. H. Flamant, J. Cuesta, and D. Bruneau, “Vertical2-μm heterodyne differential absorption lidar measurementsof mean CO2 mixing ratio in the troposphere,” under reviewJ. Atmos. Ocean. Tech. (in press).

17. J. Y. Beyon and G. J. Koch, “Energy estimation technique fordifferential absorption lidar under minimum mean squareerror criteria,” manuscript to be submitted to Opt. Eng.

18. L. Regalia-Jarlot, V. Zeninari, B. Parvitte, A. Grossel, X. Tho-mas, P. von der Heyden, and G. Durry, “A complete study of theline intensities of four bands of CO2 around 1.6 and 2:0 μm: a


comparison between Fourier transform and diode laser mea-surements,” J. Quant. Spectrosc. Radiat. Transfer 101, 325–335 (2006).

19. Y. Bai, J. Yu, M. Petros, P. Petzar, B. Trieu, H. Lee, andU. Singh, “Highly efficient Q-switched Ho:YLF laser pumpedby Tm:fiber laser,” in Proceedings of 2007 Conference onLasers and Electro-Optics, Baltimore, Md. , paper CTuN5.

20. S. A. Vay, B. E. Anderson, T. J. Conway, G. W. Sachse, J. E.Collins Jr., D. R. Blake, and D. J. Westberg, “Airborne obser-vations of the tropospheric CO2 distribution and its control-ling factors over the South Pacific basin,” J. Geophys. Res.104, 5663–5676 (1999).

21. S. A. Vay, J.-H. Woo, B. E. Anderson, K. L. Thornhill, D. R.Blake, D. J. Westberg, C. M. Kiley, M. A. Avery, G. W. Sachse,D. G. Streets, Y. Tsutsumi, and S. R. Nolf, “Influence of regio-

nal-scale anthropogenic emissions on CO2 distributions overthe western North Pacific,” J. Geophys. Res. 108, 8801 (2003).

22. B. W. Berger, K. J. Davis, C. Yi, P. S. Bakwin, and C. L. Zhao,“Long-term carbon dioxide fluxes from a very tall tower in anorthern forest: flux measurement methodology,” J. Atmos.Ocean. Tech. 18, 529–542 (2001).

23. K. J. Davis, P. S. Bakwin, C. Yi, B. W. Berger, C. Zhao, R. M.Teclaw, and J. G. Isebrands, “The annual cycles of CO2 andH2O exchange over a northern mixed forest as observed froma very tall tower,” Glob. Chang. Biol. 9, 1278–1293 (2003).

24. T. F. Refaat, S. Ismail, T. L. Mack, M. N. Abedin, S. D. Mayor,S. M. Spuler, and U. N. Singh, “Infrared phototransistor vali-dation for atmospheric remote sensing application using theRaman-shifted eye-safe aerosol lidar,” Opt. Eng. 48, 086001(2007).


side-line tunable laser transmitter for differential absorption lidar measurements of co_2: design...

Documents